synchronizing finite automata - lecture ii. automata with...
TRANSCRIPT
![Page 1: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/1.jpg)
Synchronizing Finite Automata
Lecture II. Automata with Zero and Aperiodic Automata
Mikhail Volkov
Ural Federal University
Mikhail Volkov Synchronizing Finite Automata
![Page 2: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/2.jpg)
1. Recap
Deterministic finite automata (DFA): A = 〈Q,Σ, δ〉.• Q the state set• Σ the input alphabet• δ : Q ×Σ → Q the transition function
Mikhail Volkov Synchronizing Finite Automata
![Page 3: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/3.jpg)
2. Algebraic Perspective
One can treat DFAs as unary algebras: each letter of the inputalphabet defines a unary operation on the state set.This allows us to apply to automata all standard algebraic notions,e.g., the notions of a subalgebra (subautomaton), a congruence,a quotient automaton.
Subautomata: if A = 〈Q,Σ, δ〉 is a DFA, and S ⊆ Q is such thatδ(s, a) ∈ S for all s ∈ S and a ∈ Σ, consider the DFAS := 〈S ,Σ, τ〉 where τ = σ|S×Σ. The latter equality means thatτ(s, a) := δ(s, a) for all s ∈ S and a ∈ Σ.
Any such DFA is said to be a subautomaton of A .
Mikhail Volkov Synchronizing Finite Automata
![Page 4: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/4.jpg)
2. Algebraic Perspective
One can treat DFAs as unary algebras: each letter of the inputalphabet defines a unary operation on the state set.This allows us to apply to automata all standard algebraic notions,e.g., the notions of a subalgebra (subautomaton), a congruence,a quotient automaton.
Subautomata: if A = 〈Q,Σ, δ〉 is a DFA, and S ⊆ Q is such thatδ(s, a) ∈ S for all s ∈ S and a ∈ Σ, consider the DFAS := 〈S ,Σ, τ〉 where τ = σ|S×Σ. The latter equality means thatτ(s, a) := δ(s, a) for all s ∈ S and a ∈ Σ.
Any such DFA is said to be a subautomaton of A .
Mikhail Volkov Synchronizing Finite Automata
![Page 5: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/5.jpg)
2. Algebraic Perspective
One can treat DFAs as unary algebras: each letter of the inputalphabet defines a unary operation on the state set.This allows us to apply to automata all standard algebraic notions,e.g., the notions of a subalgebra (subautomaton), a congruence,a quotient automaton.
Subautomata: if A = 〈Q,Σ, δ〉 is a DFA, and S ⊆ Q is such thatδ(s, a) ∈ S for all s ∈ S and a ∈ Σ, consider the DFAS := 〈S ,Σ, τ〉 where τ = σ|S×Σ. The latter equality means thatτ(s, a) := δ(s, a) for all s ∈ S and a ∈ Σ.
Any such DFA is said to be a subautomaton of A .
Mikhail Volkov Synchronizing Finite Automata
![Page 6: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/6.jpg)
2. Algebraic Perspective
One can treat DFAs as unary algebras: each letter of the inputalphabet defines a unary operation on the state set.This allows us to apply to automata all standard algebraic notions,e.g., the notions of a subalgebra (subautomaton), a congruence,a quotient automaton.
Subautomata: if A = 〈Q,Σ, δ〉 is a DFA, and S ⊆ Q is such thatδ(s, a) ∈ S for all s ∈ S and a ∈ Σ, consider the DFAS := 〈S ,Σ, τ〉 where τ = σ|S×Σ. The latter equality means thatτ(s, a) := δ(s, a) for all s ∈ S and a ∈ Σ.
Any such DFA is said to be a subautomaton of A .
Mikhail Volkov Synchronizing Finite Automata
![Page 7: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/7.jpg)
2. Algebraic Perspective
One can treat DFAs as unary algebras: each letter of the inputalphabet defines a unary operation on the state set.This allows us to apply to automata all standard algebraic notions,e.g., the notions of a subalgebra (subautomaton), a congruence,a quotient automaton.
Subautomata: if A = 〈Q,Σ, δ〉 is a DFA, and S ⊆ Q is such thatδ(s, a) ∈ S for all s ∈ S and a ∈ Σ, consider the DFAS := 〈S ,Σ, τ〉 where τ = σ|S×Σ. The latter equality means thatτ(s, a) := δ(s, a) for all s ∈ S and a ∈ Σ.
Any such DFA is said to be a subautomaton of A .
Mikhail Volkov Synchronizing Finite Automata
![Page 8: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/8.jpg)
3. Subautomata
A = 〈Q,Σ, δ〉
Q S
Mikhail Volkov Synchronizing Finite Automata
![Page 9: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/9.jpg)
3. Subautomata
S = 〈S ,Σ, τ〉
S
Mikhail Volkov Synchronizing Finite Automata
![Page 10: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/10.jpg)
3. Subautomata
S = 〈S ,Σ, τ〉
S
Observation: a DFA has no proper subautomata iff it is stronglyconnected.
Mikhail Volkov Synchronizing Finite Automata
![Page 11: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/11.jpg)
4. Automata with Zero
A singleton subautomaton is normally called a sink stateor just a sink. At a sink state each letter must have a loop.We study synchronizing automata and, clearly, a synchronizingautomaton may have at most one sink.If a DFA has a unique sink state, this state is called a zero stateor just a zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 12: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/12.jpg)
4. Automata with Zero
A singleton subautomaton is normally called a sink stateor just a sink. At a sink state each letter must have a loop.We study synchronizing automata and, clearly, a synchronizingautomaton may have at most one sink.If a DFA has a unique sink state, this state is called a zero stateor just a zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 13: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/13.jpg)
4. Automata with Zero
A singleton subautomaton is normally called a sink stateor just a sink. At a sink state each letter must have a loop.We study synchronizing automata and, clearly, a synchronizingautomaton may have at most one sink.If a DFA has a unique sink state, this state is called a zero stateor just a zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 14: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/14.jpg)
4. Automata with Zero
A singleton subautomaton is normally called a sink stateor just a sink. At a sink state each letter must have a loop.We study synchronizing automata and, clearly, a synchronizingautomaton may have at most one sink.If a DFA has a unique sink state, this state is called a zero stateor just a zero.
1 2 3 4 5 6
0
a
a
aaa
b b b b bb
a
a, b
Mikhail Volkov Synchronizing Finite Automata
![Page 15: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/15.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
Mikhail Volkov Synchronizing Finite Automata
![Page 16: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/16.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
Mikhail Volkov Synchronizing Finite Automata
![Page 17: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/17.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
1
3
2
4
a a
a a
bb
bb
Mikhail Volkov Synchronizing Finite Automata
![Page 18: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/18.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
1
3
2
4
a a
a a
bb
bb
π
Mikhail Volkov Synchronizing Finite Automata
![Page 19: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/19.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
1
3
2
4
a a
a a
bb
bb
π
Mikhail Volkov Synchronizing Finite Automata
![Page 20: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/20.jpg)
5. Congruences and Quotient Automata
An equivalence π on the state set Q of a DFA A = 〈Q,Σ, δ〉 iscalled a congruence if (p, q) ∈ π implies
(
δ(p, a), δ(q, a))
∈ π forall p, q ∈ Q and all a ∈ Σ. For π being a congruence, [q]π is theπ-class containing the state q.The quotient A /π is the DFA 〈Q/π,Σ, δπ〉 whereQ/π = {[q]π | q ∈ Q} and the function δπ is defined by the ruleδπ([q]π, a) = [δ(q, a)]π .
1
3
2
4
a a
a a
bb
bb
π1,2
3,4
b
a
a b
Mikhail Volkov Synchronizing Finite Automata
![Page 21: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/21.jpg)
6. Rees Congruences
Suppose that A = 〈Q,Σ, δ〉 is a DFA and S = 〈S ,Σ, τ〉 is asubautomaton of A .The partition of Q into classes one of which is S and all others aresingletons is a congruence of A .It is called the Rees congruence corresponding to S and isdenoted by ρS .Clearly, in the quotient automaton A /ρS the state S is a sink.
Mikhail Volkov Synchronizing Finite Automata
![Page 22: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/22.jpg)
6. Rees Congruences
Suppose that A = 〈Q,Σ, δ〉 is a DFA and S = 〈S ,Σ, τ〉 is asubautomaton of A .The partition of Q into classes one of which is S and all others aresingletons is a congruence of A .It is called the Rees congruence corresponding to S and isdenoted by ρS .Clearly, in the quotient automaton A /ρS the state S is a sink.
Mikhail Volkov Synchronizing Finite Automata
![Page 23: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/23.jpg)
6. Rees Congruences
Suppose that A = 〈Q,Σ, δ〉 is a DFA and S = 〈S ,Σ, τ〉 is asubautomaton of A .The partition of Q into classes one of which is S and all others aresingletons is a congruence of A .It is called the Rees congruence corresponding to S and isdenoted by ρS .Clearly, in the quotient automaton A /ρS the state S is a sink.
Mikhail Volkov Synchronizing Finite Automata
![Page 24: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/24.jpg)
6. Rees Congruences
Suppose that A = 〈Q,Σ, δ〉 is a DFA and S = 〈S ,Σ, τ〉 is asubautomaton of A .The partition of Q into classes one of which is S and all others aresingletons is a congruence of A .It is called the Rees congruence corresponding to S and isdenoted by ρS .Clearly, in the quotient automaton A /ρS the state S is a sink.
Mikhail Volkov Synchronizing Finite Automata
![Page 25: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/25.jpg)
7. Useful Observation
1. Any subautomaton of a synchronizing automaton issynchronizing, and every reset word for an automaton also servesas a reset word for any of its subautomata.
2. Any quotient of a synchronizing automaton is synchronizing,and every reset word for an automaton also serves as a reset wordfor any of its quotients.
Mikhail Volkov Synchronizing Finite Automata
![Page 26: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/26.jpg)
7. Useful Observation
1. Any subautomaton of a synchronizing automaton issynchronizing, and every reset word for an automaton also servesas a reset word for any of its subautomata.
2. Any quotient of a synchronizing automaton is synchronizing,and every reset word for an automaton also serves as a reset wordfor any of its quotients.
Mikhail Volkov Synchronizing Finite Automata
![Page 27: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/27.jpg)
8. A Reduction
Let C be any class of automata closed under taking subautomataand quotients, and let Cn stand for the class of all automata withn states in C.Let f : Z+ → N be any function such that
f (n) ≥ f (n −m + 1) + f (m) whenever n ≥ m ≥ 1.
If each synchronizing automaton in Cn which either is stronglyconnected or possesses a zero has a reset word of length f (n),then the same holds true for all synchronizing automata in Cn.
Mikhail Volkov Synchronizing Finite Automata
![Page 28: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/28.jpg)
8. A Reduction
Let C be any class of automata closed under taking subautomataand quotients, and let Cn stand for the class of all automata withn states in C.Let f : Z+ → N be any function such that
f (n) ≥ f (n −m + 1) + f (m) whenever n ≥ m ≥ 1.
If each synchronizing automaton in Cn which either is stronglyconnected or possesses a zero has a reset word of length f (n),then the same holds true for all synchronizing automata in Cn.
Mikhail Volkov Synchronizing Finite Automata
![Page 29: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/29.jpg)
8. A Reduction
Let C be any class of automata closed under taking subautomataand quotients, and let Cn stand for the class of all automata withn states in C.Let f : Z+ → N be any function such that
f (n) ≥ f (n −m + 1) + f (m) whenever n ≥ m ≥ 1.
If each synchronizing automaton in Cn which either is stronglyconnected or possesses a zero has a reset word of length f (n),then the same holds true for all synchronizing automata in Cn.
Mikhail Volkov Synchronizing Finite Automata
![Page 30: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/30.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 31: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/31.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 32: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/32.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 33: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/33.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 34: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/34.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 35: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/35.jpg)
9. A Reduction: Proof
Let A = 〈Q,Σ, δ〉 be a synchronizing automaton in Cn.Consider the set S of all states to which the automaton A can bereset and let m = |S |.If q ∈ S , then there exists a reset word w ∈ Σ∗ such thatQ.w = {q}.Then wa also is a reset word and Q.wa = {δ(q, a)} whenceδ(q, a) ∈ S .This means that, restricting the transition function δ to S × Σ, weget a subautomaton S with the state set S .Since S is synchronizing and strongly connected and since theclass C is closed under taking subautomata, we have S ∈ C.Hence, S has a reset word v of length f (m).
Mikhail Volkov Synchronizing Finite Automata
![Page 36: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/36.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 37: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/37.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 38: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/38.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 39: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/39.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 40: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/40.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 41: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/41.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 42: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/42.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 43: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/43.jpg)
10. A Reduction: End of the Proof
Now consider the Rees congruence ρS of the automaton A .The quotient A /ρS is synchronizing, has S as a zero, and hasn −m + 1 states.Since the class C is closed under taking quotients, we haveA /ρS ∈ C.Hence A /ρS has a reset word u of length f (n −m + 1).Since Q.u ⊆ S and S .v is a singleton, we conclude that alsoQ.uv ⊆ S .v is a singleton.Thus, uv is reset word for A , and the length of this word does notexceed f (n −m + 1) + f (m) ≤ f (n) according to the conditionimposed on the function f .
It is easy to check that the function f (n) = (n − 1)2 satisfies thecondition above.We see that it suffices to prove the Cerny conjecture for stronglyconnected automata and for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 44: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/44.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 45: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/45.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 46: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/46.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 47: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/47.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 48: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/48.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 49: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/49.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
a
b
b
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 50: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/50.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 51: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/51.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
a
b
babb
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 52: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/52.jpg)
11. Automata with Zero
If a synchronizing automaton with k states has a zero, then it hasa reset word of length ≤ k(k−1)
2 .
0a, b
a
a
a
b
b
b
ba a b
b
a
The algorithm makes at most k − 1 steps and the length of thesegment added in the step when t states still hold coins(k − 1 ≥ t ≥ 1) is at most k − t. The total length is
≤ 1 + 2 + · · · + (k − 1) = k(k−1)2 ≤ (k − 1)2.
Thus, the Cerny conjecture holds true for automata with zero.
Mikhail Volkov Synchronizing Finite Automata
![Page 53: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/53.jpg)
12. Rystsov’s Automata
Igor Rystsov (1997) observed that the above bound is tight.The above automaton Rn with zero has n states and n − 1 letters,is synchronizing but cannot be reset by any word of length less thann(n−1)
2 . Here the input alphabet grows with the number of states.
The problem of determining the reset threshold of n-statesynchronizing automata with zero over a fixed input alphabet stillremains open.
Mikhail Volkov Synchronizing Finite Automata
![Page 54: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/54.jpg)
12. Rystsov’s Automata
Igor Rystsov (1997) observed that the above bound is tight.
0 1 2 3 q q q n−2 n−1a1 a2 a3 an−1
Σ Σ\{a1, a2} Σ\{a2, a3} Σ\{a3 , a4} Σ\{an−2, an−1} Σ\{an−1}
The above automaton Rn with zero has n states and n − 1 letters,is synchronizing but cannot be reset by any word of length less thann(n−1)
2 . Here the input alphabet grows with the number of states.
The problem of determining the reset threshold of n-statesynchronizing automata with zero over a fixed input alphabet stillremains open.
Mikhail Volkov Synchronizing Finite Automata
![Page 55: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/55.jpg)
12. Rystsov’s Automata
Igor Rystsov (1997) observed that the above bound is tight.
0 1 2 3 q q q n−2 n−1a1 a2 a3 an−1
Σ Σ\{a1, a2} Σ\{a2, a3} Σ\{a3 , a4} Σ\{an−2, an−1} Σ\{an−1}
The above automaton Rn with zero has n states and n − 1 letters,is synchronizing but cannot be reset by any word of length less thann(n−1)
2 . Here the input alphabet grows with the number of states.
The problem of determining the reset threshold of n-statesynchronizing automata with zero over a fixed input alphabet stillremains open.
Mikhail Volkov Synchronizing Finite Automata
![Page 56: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/56.jpg)
12. Rystsov’s Automata
Igor Rystsov (1997) observed that the above bound is tight.
0 1 2 3 q q q n−2 n−1a1 a2 a3 an−1
Σ Σ\{a1, a2} Σ\{a2, a3} Σ\{a3 , a4} Σ\{an−2, an−1} Σ\{an−1}
The above automaton Rn with zero has n states and n − 1 letters,is synchronizing but cannot be reset by any word of length less thann(n−1)
2 . Here the input alphabet grows with the number of states.
The problem of determining the reset threshold of n-statesynchronizing automata with zero over a fixed input alphabet stillremains open.
Mikhail Volkov Synchronizing Finite Automata
![Page 57: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/57.jpg)
12. Rystsov’s Automata
Igor Rystsov (1997) observed that the above bound is tight.
0 1 2 3 q q q n−2 n−1a1 a2 a3 an−1
Σ Σ\{a1, a2} Σ\{a2, a3} Σ\{a3 , a4} Σ\{an−2, an−1} Σ\{an−1}
The above automaton Rn with zero has n states and n − 1 letters,is synchronizing but cannot be reset by any word of length less thann(n−1)
2 . Here the input alphabet grows with the number of states.
The problem of determining the reset threshold of n-statesynchronizing automata with zero over a fixed input alphabet stillremains open.
Mikhail Volkov Synchronizing Finite Automata
![Page 58: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/58.jpg)
13. Binary Automata with Zero
The best lower bound for the reset threshold of binary n-statesynchronizing automata with zero that holds for any sufficientlylarge n has been found by Pavel Martyugin.In 2008, he found, for every n ≥ 8, a binary n-state synchronizingautomata with zero whose reset threshold is
⌈
14n
2 + 32n − 4
⌉
.
Very recently, Dmitry Ananichev and Vojtech Vorel haveconstructed a new infinite series of binary synchronizing automatawith zero.Automata in this series have n ≥ 16 states, n ≡ 4 (mod 12), andthe reset threshold of the n-state automaton in the series equals14n
2 + 2n − 9.
Mikhail Volkov Synchronizing Finite Automata
![Page 59: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/59.jpg)
13. Binary Automata with Zero
The best lower bound for the reset threshold of binary n-statesynchronizing automata with zero that holds for any sufficientlylarge n has been found by Pavel Martyugin.In 2008, he found, for every n ≥ 8, a binary n-state synchronizingautomata with zero whose reset threshold is
⌈
14n
2 + 32n − 4
⌉
.
Very recently, Dmitry Ananichev and Vojtech Vorel haveconstructed a new infinite series of binary synchronizing automatawith zero.Automata in this series have n ≥ 16 states, n ≡ 4 (mod 12), andthe reset threshold of the n-state automaton in the series equals14n
2 + 2n − 9.
Mikhail Volkov Synchronizing Finite Automata
![Page 60: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/60.jpg)
13. Binary Automata with Zero
The best lower bound for the reset threshold of binary n-statesynchronizing automata with zero that holds for any sufficientlylarge n has been found by Pavel Martyugin.In 2008, he found, for every n ≥ 8, a binary n-state synchronizingautomata with zero whose reset threshold is
⌈
14n
2 + 32n − 4
⌉
.
Very recently, Dmitry Ananichev and Vojtech Vorel haveconstructed a new infinite series of binary synchronizing automatawith zero.Automata in this series have n ≥ 16 states, n ≡ 4 (mod 12), andthe reset threshold of the n-state automaton in the series equals14n
2 + 2n − 9.
Mikhail Volkov Synchronizing Finite Automata
![Page 61: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/61.jpg)
13. Binary Automata with Zero
The best lower bound for the reset threshold of binary n-statesynchronizing automata with zero that holds for any sufficientlylarge n has been found by Pavel Martyugin.In 2008, he found, for every n ≥ 8, a binary n-state synchronizingautomata with zero whose reset threshold is
⌈
14n
2 + 32n − 4
⌉
.
Very recently, Dmitry Ananichev and Vojtech Vorel haveconstructed a new infinite series of binary synchronizing automatawith zero.Automata in this series have n ≥ 16 states, n ≡ 4 (mod 12), andthe reset threshold of the n-state automaton in the series equals14n
2 + 2n − 9.
Mikhail Volkov Synchronizing Finite Automata
![Page 62: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/62.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
Mikhail Volkov Synchronizing Finite Automata
![Page 63: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/63.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
Mikhail Volkov Synchronizing Finite Automata
![Page 64: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/64.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
Mikhail Volkov Synchronizing Finite Automata
![Page 65: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/65.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
Mikhail Volkov Synchronizing Finite Automata
![Page 66: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/66.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
q w w . . . wwk , k > 1
Mikhail Volkov Synchronizing Finite Automata
![Page 67: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/67.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
q w w . . . wwk , k > 1
Mikhail Volkov Synchronizing Finite Automata
![Page 68: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/68.jpg)
14. Aperiodic Automata
The transition monoid of a DFA A = 〈Q,Σ, δ〉 consists of alltransformations δ(xy,w) : Q → Q induced by words w ∈ Σ∗.
A monoid is said to be aperiodic if all its subgroups are singletons.
A DFA is called aperiodic (or counter-free) if its transition monoidis aperiodic.
An equivalent ‘elementary’ formulation: A = 〈Q,Σ, δ〉 is aperiodiciff for every q ∈ Q and every w ∈ Σ∗ there exists a positive integerm such that q .wm = q .wm+1.
q w w . . . ww
Mikhail Volkov Synchronizing Finite Automata
![Page 69: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/69.jpg)
15. Examples
The Cerny automaton C4 is not aperiodic since the letter b acts asa cyclic permutation of the states and thus generates a 4-elementsubgroup in the transition monoid of C4.
0 1
23
a
b
b
b
b
a
aa
The following automaton is aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 70: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/70.jpg)
15. Examples
The Cerny automaton C4 is not aperiodic since the letter b acts asa cyclic permutation of the states and thus generates a 4-elementsubgroup in the transition monoid of C4.The following automaton is aperiodic:
1
3
2
4
a a
a a
bb
bb
Mikhail Volkov Synchronizing Finite Automata
![Page 71: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/71.jpg)
15. Examples
The Cerny automaton C4 is not aperiodic since the letter b acts asa cyclic permutation of the states and thus generates a 4-elementsubgroup in the transition monoid of C4.The following automaton is aperiodic:
1
3
2
4
a a
a a
bb
bb
1 2 3 4
a 3 4 3 4b 1 1 2 2
Mikhail Volkov Synchronizing Finite Automata
![Page 72: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/72.jpg)
15. Examples
The Cerny automaton C4 is not aperiodic since the letter b acts asa cyclic permutation of the states and thus generates a 4-elementsubgroup in the transition monoid of C4.The following automaton is aperiodic:
1
3
2
4
a a
a a
bb
bb
1 2 3 4
a 3 4 3 4b 1 1 2 2ab 2 2 2 2ba 3 3 4 4b2 1 1 1 1
a2 = a
Mikhail Volkov Synchronizing Finite Automata
![Page 73: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/73.jpg)
15. Examples
The Cerny automaton C4 is not aperiodic since the letter b acts asa cyclic permutation of the states and thus generates a 4-elementsubgroup in the transition monoid of C4.The following automaton is aperiodic:
1
3
2
4
a a
a a
bb
bb
1 2 3 4
a 3 4 3 4b 1 1 2 2ab 2 2 2 2ba 3 3 4 4b2 1 1 1 1aba 4 4 4 4b2a 3 3 3 3
a2 = a, ab2 = b2, bab = ab, b3 = b2
Mikhail Volkov Synchronizing Finite Automata
![Page 74: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/74.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 75: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/75.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 76: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/76.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 77: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/77.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 78: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/78.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 79: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/79.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
Mikhail Volkov Synchronizing Finite Automata
![Page 80: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/80.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
w
w
w
w
w
Mikhail Volkov Synchronizing Finite Automata
![Page 81: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/81.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
q .w q<
Mikhail Volkov Synchronizing Finite Automata
![Page 82: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/82.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
q .w2
q .w q<
<
Mikhail Volkov Synchronizing Finite Automata
![Page 83: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/83.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
q .w3
q .w2
q .w q
<
<<
Mikhail Volkov Synchronizing Finite Automata
![Page 84: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/84.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
q .w3
q .w2 q .w4
q .w q
< <
<<
Mikhail Volkov Synchronizing Finite Automata
![Page 85: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/85.jpg)
16. Complexity
In general, there is no way to verify whether or not a given DFAA = 〈Q,Σ, δ〉 is aperiodic avoiding the calculation of its transitionmonoid and the cardinality of the monoid can reach |Q||Q|. Theproblem is known to be PSPACE-complete (Sang Cho and DungT. Huynh, Theor. Comput. Sci. 88 (1991) 99–116).In some cases, however, aperiodicity is granted.A DFA A = 〈Q,Σ, δ〉 is monotonic if Q admits a linear order ≤such that, for each a ∈ Σ, the transformation δ(xy, a) preserves ≤:p ≤ q ⇒ δ(p, a) ≤ δ(q, a). Monotonic automata are aperiodic:
q .w3
q .w2 q .w4
q .w q
<< – contradiction!
<
<
<
Mikhail Volkov Synchronizing Finite Automata
![Page 86: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/86.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 87: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/87.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 88: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/88.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 89: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/89.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 90: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/90.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 91: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/91.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 92: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/92.jpg)
17. Generalized Monotonicity
A binary relation ρ on the state set of a DFA A = 〈Q,Σ, δ〉 isstable if (p, q) ∈ ρ implies
(
δ(p, a), δ(q, a))
∈ ρ for all p, q ∈ Q
and a ∈ Σ.We call a DFA A generalized monotonic of level ℓ if it admitsa strictly increasing chain of stable binary relationsρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓ, satisfying the following conditions:• ρ0 is the equality;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a linear order on eachπi/πi−1-class;• πℓ is the universal relation.
Monotonic automata are precisely generalized monotonic automataof level 1.The aperiodic automaton in our example is a generalizedmonotonic automaton of level 2.
Mikhail Volkov Synchronizing Finite Automata
![Page 93: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/93.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 94: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/94.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 95: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/95.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 96: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/96.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 97: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/97.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 98: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/98.jpg)
18. Example Revisited
1
3
2
4
a a
a a
bb
bb
π1
1,2
3,4
b
a
a b
Endowing Q with the order ≤1 such that 1 <1 2 and 3 <1 4, weget a linear order on each π1-class. If we order Q/π1 by letting{1, 2} <2 {3, 4}, the quotient automaton becomes monotonic.It can be shown that the automaton is not monotonic. Moreover,it cannot be emulated by any monotonic automaton.
In fact, the hierarchy of generalized monotonic automata is strict:there are automata of each level ℓ = 1, 2, . . . , and everygeneralized monotonic automaton is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 99: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/99.jpg)
19. Why Aperiodic Automata?
The importance of aperiodic automata was understood at thebeginning of 1960s. As usual, it has 3 sources and 3 components:It is remarkable that each of these directions has led to a majoropen problem, and the 3 problems play nowadays a central rolein the theory of finite automata.
Mikhail Volkov Synchronizing Finite Automata
![Page 100: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/100.jpg)
19. Why Aperiodic Automata?
The importance of aperiodic automata was understood at thebeginning of 1960s. As usual, it has 3 sources and 3 components:• Star-free regular expressions (Schutzenberger);It is remarkable that each of these directions has led to a majoropen problem, and the 3 problems play nowadays a central rolein the theory of finite automata.
Mikhail Volkov Synchronizing Finite Automata
![Page 101: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/101.jpg)
19. Why Aperiodic Automata?
The importance of aperiodic automata was understood at thebeginning of 1960s. As usual, it has 3 sources and 3 components:• Star-free regular expressions (Schutzenberger);• Krohn–Rhodes decompositions of finite automata;It is remarkable that each of these directions has led to a majoropen problem, and the 3 problems play nowadays a central rolein the theory of finite automata.
Mikhail Volkov Synchronizing Finite Automata
![Page 102: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/102.jpg)
19. Why Aperiodic Automata?
The importance of aperiodic automata was understood at thebeginning of 1960s. As usual, it has 3 sources and 3 components:• Star-free regular expressions (Schutzenberger);• Krohn–Rhodes decompositions of finite automata;• Logical characterizations of regular languages (McNaughton).It is remarkable that each of these directions has led to a majoropen problem, and the 3 problems play nowadays a central rolein the theory of finite automata.
Mikhail Volkov Synchronizing Finite Automata
![Page 103: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/103.jpg)
19. Why Aperiodic Automata?
The importance of aperiodic automata was understood at thebeginning of 1960s. As usual, it has 3 sources and 3 components:• Star-free regular expressions (Schutzenberger);• Krohn–Rhodes decompositions of finite automata;• Logical characterizations of regular languages (McNaughton).It is remarkable that each of these directions has led to a majoropen problem, and the 3 problems play nowadays a central rolein the theory of finite automata.
Mikhail Volkov Synchronizing Finite Automata
![Page 104: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/104.jpg)
20. Star-free Regular Expressions
By Kleene’s theorem every regular language can be described bya regular expression, say, ((a + ba)∗ab)∗(b + aa)∗. Here wordsdenote corresponding singleton languages, + stands for union,concatenation means product and ∗ is the Kleene star (iteration).
The Kleene star is clearly the most ‘infinite’ operation. One cannoteliminate it because neither union nor product can produce infinitelanguages from finite ones. However, one can use also complement(the class of regular languages is closed under complement byKleene’s theorem). An extended regular expression is built fromwords by using union, product, Kleene star, and complement, say,((a + ba)Cab)∗(b + (aa)C )∗.
The complement of a finite language is infinite. Can one get rid ofthe Kleene star in this setting?
Mikhail Volkov Synchronizing Finite Automata
![Page 105: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/105.jpg)
20. Star-free Regular Expressions
By Kleene’s theorem every regular language can be described bya regular expression, say, ((a + ba)∗ab)∗(b + aa)∗. Here wordsdenote corresponding singleton languages, + stands for union,concatenation means product and ∗ is the Kleene star (iteration).
The Kleene star is clearly the most ‘infinite’ operation. One cannoteliminate it because neither union nor product can produce infinitelanguages from finite ones. However, one can use also complement(the class of regular languages is closed under complement byKleene’s theorem). An extended regular expression is built fromwords by using union, product, Kleene star, and complement, say,((a + ba)Cab)∗(b + (aa)C )∗.
The complement of a finite language is infinite. Can one get rid ofthe Kleene star in this setting?
Mikhail Volkov Synchronizing Finite Automata
![Page 106: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/106.jpg)
20. Star-free Regular Expressions
By Kleene’s theorem every regular language can be described bya regular expression, say, ((a + ba)∗ab)∗(b + aa)∗. Here wordsdenote corresponding singleton languages, + stands for union,concatenation means product and ∗ is the Kleene star (iteration).
The Kleene star is clearly the most ‘infinite’ operation. One cannoteliminate it because neither union nor product can produce infinitelanguages from finite ones. However, one can use also complement(the class of regular languages is closed under complement byKleene’s theorem). An extended regular expression is built fromwords by using union, product, Kleene star, and complement, say,((a + ba)Cab)∗(b + (aa)C )∗.
The complement of a finite language is infinite. Can one get rid ofthe Kleene star in this setting?
Mikhail Volkov Synchronizing Finite Automata
![Page 107: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/107.jpg)
20. Star-free Regular Expressions
By Kleene’s theorem every regular language can be described bya regular expression, say, ((a + ba)∗ab)∗(b + aa)∗. Here wordsdenote corresponding singleton languages, + stands for union,concatenation means product and ∗ is the Kleene star (iteration).
The Kleene star is clearly the most ‘infinite’ operation. One cannoteliminate it because neither union nor product can produce infinitelanguages from finite ones. However, one can use also complement(the class of regular languages is closed under complement byKleene’s theorem). An extended regular expression is built fromwords by using union, product, Kleene star, and complement, say,((a + ba)Cab)∗(b + (aa)C )∗.
The complement of a finite language is infinite. Can one get rid ofthe Kleene star in this setting?
Mikhail Volkov Synchronizing Finite Automata
![Page 108: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/108.jpg)
20. Star-free Regular Expressions
By Kleene’s theorem every regular language can be described bya regular expression, say, ((a + ba)∗ab)∗(b + aa)∗. Here wordsdenote corresponding singleton languages, + stands for union,concatenation means product and ∗ is the Kleene star (iteration).
The Kleene star is clearly the most ‘infinite’ operation. One cannoteliminate it because neither union nor product can produce infinitelanguages from finite ones. However, one can use also complement(the class of regular languages is closed under complement byKleene’s theorem). An extended regular expression is built fromwords by using union, product, Kleene star, and complement, say,((a + ba)Cab)∗(b + (aa)C )∗.
The complement of a finite language is infinite. Can one get rid ofthe Kleene star in this setting?
Mikhail Volkov Synchronizing Finite Automata
![Page 109: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/109.jpg)
21. Star-free Regular Expressions
In some cases we can:
(ab)∗
Here E1 \ E2 = E1 ∩ EC2 can be expressed as (EC
1 + E2)C
by De Morgan’s law.To understand the formula, observe that a+ aC = Σ∗.
However, for the language (a2)∗ that looks alike (ab)∗ we wouldnot be able to construct a star-free extended regular expression.
How can one distinguish between regular languages that need starand ‘star-free’ languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 110: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/110.jpg)
21. Star-free Regular Expressions
In some cases we can:
(ab)∗ = ε+a(a+aC )b\(
(a+aC )aa(a+aC )+(a+aC )bb(a+aC ))
.
Here E1 \ E2 = E1 ∩ EC2 can be expressed as (EC
1 + E2)C
by De Morgan’s law.To understand the formula, observe that a+ aC = Σ∗.
However, for the language (a2)∗ that looks alike (ab)∗ we wouldnot be able to construct a star-free extended regular expression.
How can one distinguish between regular languages that need starand ‘star-free’ languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 111: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/111.jpg)
21. Star-free Regular Expressions
In some cases we can:
(ab)∗ = ε+a(a+aC )b\(
(a+aC )aa(a+aC )+(a+aC )bb(a+aC ))
.
Here E1 \ E2 = E1 ∩ EC2 can be expressed as (EC
1 + E2)C
by De Morgan’s law.To understand the formula, observe that a+ aC = Σ∗.
However, for the language (a2)∗ that looks alike (ab)∗ we wouldnot be able to construct a star-free extended regular expression.
How can one distinguish between regular languages that need starand ‘star-free’ languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 112: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/112.jpg)
21. Star-free Regular Expressions
In some cases we can:
(ab)∗ = ε+a(a+aC )b\(
(a+aC )aa(a+aC )+(a+aC )bb(a+aC ))
.
Here E1 \ E2 = E1 ∩ EC2 can be expressed as (EC
1 + E2)C
by De Morgan’s law.To understand the formula, observe that a+ aC = Σ∗.
However, for the language (a2)∗ that looks alike (ab)∗ we wouldnot be able to construct a star-free extended regular expression.
How can one distinguish between regular languages that need starand ‘star-free’ languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 113: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/113.jpg)
21. Star-free Regular Expressions
In some cases we can:
(ab)∗ = ε+a(a+aC )b\(
(a+aC )aa(a+aC )+(a+aC )bb(a+aC ))
.
Here E1 \ E2 = E1 ∩ EC2 can be expressed as (EC
1 + E2)C
by De Morgan’s law.To understand the formula, observe that a+ aC = Σ∗.
However, for the language (a2)∗ that looks alike (ab)∗ we wouldnot be able to construct a star-free extended regular expression.
How can one distinguish between regular languages that need starand ‘star-free’ languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 114: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/114.jpg)
22. Schutzenberger’s Theorem
Schutzenberger’s Theorem, 1964
A regular language L admits a star-free extended regular expressioniff the minimal automaton of L is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 115: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/115.jpg)
22. Schutzenberger’s Theorem
Schutzenberger’s Theorem, 1964
A regular language L admits a star-free extended regular expressioniff the minimal automaton of L is aperiodic.
For instance, for (ab)∗ and (a2)∗ the minimal automata are
a
bab
a, b
and
a
a
Mikhail Volkov Synchronizing Finite Automata
![Page 116: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/116.jpg)
23. Extended Star Height Problem
By the (extended) star height of a regular language L we meanthe minimum number of nested stars over all (extended) regularexpressions representing L.It is known that there exist regular languages of any given starheight and that, given a language, its star height can be decided.However analogous problems are open for extended star height.
Mikhail Volkov Synchronizing Finite Automata
![Page 117: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/117.jpg)
23. Extended Star Height Problem
By the (extended) star height of a regular language L we meanthe minimum number of nested stars over all (extended) regularexpressions representing L.It is known that there exist regular languages of any given starheight and that, given a language, its star height can be decided.However analogous problems are open for extended star height.
Mikhail Volkov Synchronizing Finite Automata
![Page 118: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/118.jpg)
23. Extended Star Height Problem
By the (extended) star height of a regular language L we meanthe minimum number of nested stars over all (extended) regularexpressions representing L.It is known that there exist regular languages of any given starheight and that, given a language, its star height can be decided.However analogous problems are open for extended star height.
Mikhail Volkov Synchronizing Finite Automata
![Page 119: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/119.jpg)
23. Extended Star Height Problem
By the (extended) star height of a regular language L we meanthe minimum number of nested stars over all (extended) regularexpressions representing L.It is known that there exist regular languages of any given starheight and that, given a language, its star height can be decided.However analogous problems are open for extended star height.
Extended Star Height Problem
Is there a regular language of extended star height > 1?Is the class of languages of extended star height 1 decidable?
Mikhail Volkov Synchronizing Finite Automata
![Page 120: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/120.jpg)
24. Krohn–Rhodes Theorem
A DFA is said to be a group automaton if every letter acts as apermutation of the state set. Group automata are just Cayleygraphs of groups and are antipodes of aperiodic automata.Thus, A decomposes into counter (=group) and non-counter(=aperiodic) components. Group components can be furtherdecomposed into cascade compositions of Cayley graphs of simplegroups while aperiodic components are cascade compositions offlip-flops and their 1-letter subautomata.
Mikhail Volkov Synchronizing Finite Automata
![Page 121: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/121.jpg)
24. Krohn–Rhodes Theorem
A DFA is said to be a group automaton if every letter acts as apermutation of the state set. Group automata are just Cayleygraphs of groups and are antipodes of aperiodic automata.Thus, A decomposes into counter (=group) and non-counter(=aperiodic) components. Group components can be furtherdecomposed into cascade compositions of Cayley graphs of simplegroups while aperiodic components are cascade compositions offlip-flops and their 1-letter subautomata.
Mikhail Volkov Synchronizing Finite Automata
![Page 122: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/122.jpg)
24. Krohn–Rhodes Theorem
A DFA is said to be a group automaton if every letter acts as apermutation of the state set. Group automata are just Cayleygraphs of groups and are antipodes of aperiodic automata.
Krohn–Rhodes Theorem, 1962
Every finite automaton A can be emulated by a cascadecomposition of an alternating sequence of aperiodic and groupautomata derived from A .
Thus, A decomposes into counter (=group) and non-counter(=aperiodic) components. Group components can be furtherdecomposed into cascade compositions of Cayley graphs of simplegroups while aperiodic components are cascade compositions offlip-flops and their 1-letter subautomata.
Mikhail Volkov Synchronizing Finite Automata
![Page 123: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/123.jpg)
24. Krohn–Rhodes Theorem
A DFA is said to be a group automaton if every letter acts as apermutation of the state set. Group automata are just Cayleygraphs of groups and are antipodes of aperiodic automata.
Krohn–Rhodes Theorem, 1962
Every finite automaton A can be emulated by a cascadecomposition of an alternating sequence of aperiodic and groupautomata derived from A .
Thus, A decomposes into counter (=group) and non-counter(=aperiodic) components. Group components can be furtherdecomposed into cascade compositions of Cayley graphs of simplegroups while aperiodic components are cascade compositions offlip-flops and their 1-letter subautomata.
Mikhail Volkov Synchronizing Finite Automata
![Page 124: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/124.jpg)
24. Krohn–Rhodes Theorem
A DFA is said to be a group automaton if every letter acts as apermutation of the state set. Group automata are just Cayleygraphs of groups and are antipodes of aperiodic automata.
Krohn–Rhodes Theorem, 1962
Every finite automaton A can be emulated by a cascadecomposition of an alternating sequence of aperiodic and groupautomata derived from A .
Thus, A decomposes into counter (=group) and non-counter(=aperiodic) components. Group components can be furtherdecomposed into cascade compositions of Cayley graphs of simplegroups while aperiodic components are cascade compositions offlip-flops and their 1-letter subautomata.
On OffOff
On
Off
On
Mikhail Volkov Synchronizing Finite Automata
![Page 125: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/125.jpg)
25. Group Complexity
The minimum number of group components in the Krohn–Rhodesdecomposition of A is called the group complexity of A . Thisparameter gives rise to an infinite hierarchy.
Mikhail Volkov Synchronizing Finite Automata
![Page 126: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/126.jpg)
25. Group Complexity
The minimum number of group components in the Krohn–Rhodesdecomposition of A is called the group complexity of A . Thisparameter gives rise to an infinite hierarchy.
Mikhail Volkov Synchronizing Finite Automata
![Page 127: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/127.jpg)
25. Group Complexity
The minimum number of group components in the Krohn–Rhodesdecomposition of A is called the group complexity of A . Thisparameter gives rise to an infinite hierarchy.
Group Complexity Problem
Given a finite automaton A , can one decide the group complexityof A ?In particular, can we decide if the group complexity of A is 1?
Mikhail Volkov Synchronizing Finite Automata
![Page 128: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/128.jpg)
26. Logic for Words
There is a magic triangle
Automata Languages
Logic
Logic for words has first order variables (positions) that take valuesin {1, 2, . . . }, second order variables (sets of positions) whosevalues are subsets of {1, 2, . . . }, the usual connectives andquantifiers, the predicate symbol < with the usual meaning(and maybe some additional numerical predicates), and a specialpredicate Qa for each letter a with the meaning: Qax is true iffthe position x holds the letter a.
Mikhail Volkov Synchronizing Finite Automata
![Page 129: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/129.jpg)
26. Logic for Words
There is a magic triangle
Automata Languages
Logic
Logic for words has first order variables (positions) that take valuesin {1, 2, . . . }, second order variables (sets of positions) whosevalues are subsets of {1, 2, . . . }, the usual connectives andquantifiers, the predicate symbol < with the usual meaning(and maybe some additional numerical predicates), and a specialpredicate Qa for each letter a with the meaning: Qax is true iffthe position x holds the letter a.
Mikhail Volkov Synchronizing Finite Automata
![Page 130: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/130.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
Ψ : ∃x(
¬(
∃y(x < y))
)
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 131: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/131.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
Ψ : ∃x(
¬(
∃y(x < y))
)
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 132: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/132.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 133: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/133.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 134: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/134.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 135: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/135.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 136: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/136.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 137: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/137.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 138: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/138.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
Here y=x+1 abbreviates (x < y)&¬(
∃z ((x < z)&(z < y)))
.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 139: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/139.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 140: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/140.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 141: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/141.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 142: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/142.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
Mikhail Volkov Synchronizing Finite Automata
![Page 143: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/143.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
This (monadic second order) formula defines the language (a2)∗.
Mikhail Volkov Synchronizing Finite Automata
![Page 144: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/144.jpg)
27. Examples
Any closed formula of this logic defines a language.
Φa : ∀x(
¬(
∃y(y < x))
→ Qax)
all words starting with a
Ψ : ∃x(
¬(
∃y(x < y))
)
all finite words
Ψb : Ψ&∀x(
¬(
∃y(x < y))
→ Qbx)
all finite words ending with b
Φa &Ψb &∀x∀y(
(y=x+1) → ((Qax → Qby)& (Qbx → Qay)))
This (first order) formula defines the language (ab)∗.
Ψ&∀x (Qax)&∃H(
∀x∀y(
(y=x+1) → ((x ∈ H) ↔ ¬(y ∈ H)))
&
∀x((
¬(∃y(y < x)) → (x ∈ H))
&(
¬(∃y(x < y)) → ¬(x ∈ H)))
)
This (monadic second order) formula defines the language (a2)∗.
Monadic second order formulas define precisely regular languages(Buchi, 1960), but we would not be able to construct a first orderformula defining (a2)∗.
Mikhail Volkov Synchronizing Finite Automata
![Page 145: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/145.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
Mikhail Volkov Synchronizing Finite Automata
![Page 146: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/146.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
Mikhail Volkov Synchronizing Finite Automata
![Page 147: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/147.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
A natural complexity measure for first order formulas is thenumber of alternations of logical quantifiers in the prenex form.
Mikhail Volkov Synchronizing Finite Automata
![Page 148: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/148.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
A natural complexity measure for first order formulas is thenumber of alternations of logical quantifiers in the prenex form.The minimum number of quantifier alternations over all first orderformulas representing a given star-free language L is called thedot-depth of L.
Mikhail Volkov Synchronizing Finite Automata
![Page 149: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/149.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
A natural complexity measure for first order formulas is thenumber of alternations of logical quantifiers in the prenex form.The minimum number of quantifier alternations over all first orderformulas representing a given star-free language L is called thedot-depth of L. This parameter gives rise to an infinite hierarchy.
Mikhail Volkov Synchronizing Finite Automata
![Page 150: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/150.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
A natural complexity measure for first order formulas is thenumber of alternations of logical quantifiers in the prenex form.The minimum number of quantifier alternations over all first orderformulas representing a given star-free language L is called thedot-depth of L. This parameter gives rise to an infinite hierarchy.
Dot-Depth Problem
Given a star-free language L, can one decide the dot-depth of L?In particular, can we decide if the dot-depth of L is 3?
Mikhail Volkov Synchronizing Finite Automata
![Page 151: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/151.jpg)
28. McNaughton’s Theorem
Can one distinguish between ‘second’ and ‘first’ order languages?
McNaughton’s Theorem, 1966
A regular language L admits a description by a first order formulaiff the minimal automaton of L is aperiodic.
A natural complexity measure for first order formulas is thenumber of alternations of logical quantifiers in the prenex form.The minimum number of quantifier alternations over all first orderformulas representing a given star-free language L is called thedot-depth of L. This parameter gives rise to an infinite hierarchy.
Dot-Depth Problem
Given a star-free language L, can one decide the dot-depth of L?In particular, can we decide if the dot-depth of L is 3?
Dot-depth 1 and dot-depth 2 are known to be decidable(Knast, 1980, for 1 and Place-Zeitoun, 2014, for 2)
Mikhail Volkov Synchronizing Finite Automata
![Page 152: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/152.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 153: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/153.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 154: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/154.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 155: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/155.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 156: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/156.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 157: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/157.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 158: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/158.jpg)
29. Back to Synchronization
Here we aim to study aperiodic automata from the viewpoint ofsynchronization (in particular, to prove the Cerny conjecture foraperiodic automata).
As discussed above, we may restrict to strongly connectedautomata.Here we encounter a small surprise: every strongly connected
aperiodic automaton A = 〈Q,Σ, δ〉 is synchronizing.
Indeed, take any q, q′ ∈ Q. Since A is strongly connected, thereexists w ∈ Σ∗ such that q .w = q′. On the other hand, A isaperiodic whence there exists a positive integer m such thatq .wm = q .wm+1. Applying wm to the equality q .w = q′, we getq .wm+1 = q′ .wm whence q .wm = q′ .wm. Thus, every pair ofstates can be synchronized, and by Cerny’s criterion, this ensuresthat A is synchronizing.
Mikhail Volkov Synchronizing Finite Automata
![Page 159: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/159.jpg)
Partial Order
Avraham Trahtman (The Cerny conjecture for aperiodic automata,Discrete Math. Theor. Comp. Sci. 9(2) (2007) 3–10) has provedthat every synchronizing aperiodic automaton with n states has areset word of length n(n−1)
2 (so less than (n − 1)2).
Recall that n(n−1)2 is precisely Rystsov’s bound for n-state
synchronizing automata having a sink. Thus, it remains to provethat every strongly connected aperiodic automaton with n stateshas a reset word of length n(n−1)
2 .
The key observation by Trahtman is that every strongly connectedaperiodic automaton admits a non-trivial stable partial order.
Mikhail Volkov Synchronizing Finite Automata
![Page 160: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/160.jpg)
Partial Order
Avraham Trahtman (The Cerny conjecture for aperiodic automata,Discrete Math. Theor. Comp. Sci. 9(2) (2007) 3–10) has provedthat every synchronizing aperiodic automaton with n states has areset word of length n(n−1)
2 (so less than (n − 1)2).
Recall that n(n−1)2 is precisely Rystsov’s bound for n-state
synchronizing automata having a sink. Thus, it remains to provethat every strongly connected aperiodic automaton with n stateshas a reset word of length n(n−1)
2 .
The key observation by Trahtman is that every strongly connectedaperiodic automaton admits a non-trivial stable partial order.
Mikhail Volkov Synchronizing Finite Automata
![Page 161: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/161.jpg)
Partial Order
Avraham Trahtman (The Cerny conjecture for aperiodic automata,Discrete Math. Theor. Comp. Sci. 9(2) (2007) 3–10) has provedthat every synchronizing aperiodic automaton with n states has areset word of length n(n−1)
2 (so less than (n − 1)2).
Recall that n(n−1)2 is precisely Rystsov’s bound for n-state
synchronizing automata having a sink. Thus, it remains to provethat every strongly connected aperiodic automaton with n stateshas a reset word of length n(n−1)
2 .
The key observation by Trahtman is that every strongly connectedaperiodic automaton admits a non-trivial stable partial order.
Mikhail Volkov Synchronizing Finite Automata
![Page 162: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/162.jpg)
Partial Order
Avraham Trahtman (The Cerny conjecture for aperiodic automata,Discrete Math. Theor. Comp. Sci. 9(2) (2007) 3–10) has provedthat every synchronizing aperiodic automaton with n states has areset word of length n(n−1)
2 (so less than (n − 1)2).
Recall that n(n−1)2 is precisely Rystsov’s bound for n-state
synchronizing automata having a sink. Thus, it remains to provethat every strongly connected aperiodic automaton with n stateshas a reset word of length n(n−1)
2 .
The key observation by Trahtman is that every strongly connectedaperiodic automaton admits a non-trivial stable partial order.
Mikhail Volkov Synchronizing Finite Automata
![Page 163: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/163.jpg)
31. Constructing Partial Order
Given a DFA A = 〈Q,Σ, δ〉, its square A [2] = 〈Q ×Q,Σ, δ[2]〉 isdefined by δ[2]
(
(q, p), a)
=(
δ(q, a), δ(p, a))
.Warning: it is not quite the same as the automaton on all at most2-element subsets that we considered in Lecture I.
If A is synchronizing and strongly connected, then A [2] hasa least strongly connected component D = {(q, q) | q ∈ Q}. LetK be a strongly connected component immediately following D inthe natural order of strongly connected components.
Mikhail Volkov Synchronizing Finite Automata
![Page 164: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/164.jpg)
31. Constructing Partial Order
Given a DFA A = 〈Q,Σ, δ〉, its square A [2] = 〈Q ×Q,Σ, δ[2]〉 isdefined by δ[2]
(
(q, p), a)
=(
δ(q, a), δ(p, a))
.Warning: it is not quite the same as the automaton on all at most2-element subsets that we considered in Lecture I.
If A is synchronizing and strongly connected, then A [2] hasa least strongly connected component D = {(q, q) | q ∈ Q}. LetK be a strongly connected component immediately following D inthe natural order of strongly connected components.
Mikhail Volkov Synchronizing Finite Automata
![Page 165: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/165.jpg)
31. Constructing Partial Order
Given a DFA A = 〈Q,Σ, δ〉, its square A [2] = 〈Q ×Q,Σ, δ[2]〉 isdefined by δ[2]
(
(q, p), a)
=(
δ(q, a), δ(p, a))
.Warning: it is not quite the same as the automaton on all at most2-element subsets that we considered in Lecture I.
If A is synchronizing and strongly connected, then A [2] hasa least strongly connected component D = {(q, q) | q ∈ Q}. LetK be a strongly connected component immediately following D inthe natural order of strongly connected components.
Mikhail Volkov Synchronizing Finite Automata
![Page 166: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/166.jpg)
31. Constructing Partial Order
Given a DFA A = 〈Q,Σ, δ〉, its square A [2] = 〈Q ×Q,Σ, δ[2]〉 isdefined by δ[2]
(
(q, p), a)
=(
δ(q, a), δ(p, a))
.Warning: it is not quite the same as the automaton on all at most2-element subsets that we considered in Lecture I.
If A is synchronizing and strongly connected, then A [2] hasa least strongly connected component D = {(q, q) | q ∈ Q}. LetK be a strongly connected component immediately following D inthe natural order of strongly connected components.
D
K
Mikhail Volkov Synchronizing Finite Automata
![Page 167: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/167.jpg)
31. Constructing Partial Order
Given a DFA A = 〈Q,Σ, δ〉, its square A [2] = 〈Q ×Q,Σ, δ[2]〉 isdefined by δ[2]
(
(q, p), a)
=(
δ(q, a), δ(p, a))
.Warning: it is not quite the same as the automaton on all at most2-element subsets that we considered in Lecture I.
If A is synchronizing and strongly connected, then A [2] hasa least strongly connected component D = {(q, q) | q ∈ Q}. LetK be a strongly connected component immediately following D inthe natural order of strongly connected components.
D
K
Then K ∪D is a non-trivial sta-ble reflexive relation on Q.
Mikhail Volkov Synchronizing Finite Automata
![Page 168: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/168.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 169: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/169.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 170: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/170.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 171: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/171.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 172: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/172.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 173: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/173.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 174: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/174.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p.
Mikhail Volkov Synchronizing Finite Automata
![Page 175: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/175.jpg)
32. A is aperiodic
Let �K be the transitive closure of K ∪ D. It is clear that �K isnon-trivial, stable, reflexive and transitive.Now we show that �K is antisymmetric whenever A is aperiodic.
Suppose that there are p, q ∈ Q such that p 6= q andp �K q �K p. Then there is a sequence of p0, p1, . . . , pk ∈ Q suchthat k > 1, p0 = p = pk , q = pj for some j , 0 < j < k , and(pi , pi+1) ∈ K for all i = 0, 1, . . . , k − 1. We choose the shortestsuch sequence p0, p1, . . . , pk (over all possible ‘obstacles’ (p, q) toantisymmetry).
If k = 2, then we have p0 = p = p2, p1 = q and (p, q), (q, p) ∈ K .By the definition of K , there exists w ∈ Σ∗ such that(p, q) .w = (q, p), that is, p .w = q, q .w = p. This clearlycontradicts the assumption that A is aperiodic:
p qw
w
Mikhail Volkov Synchronizing Finite Automata
![Page 176: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/176.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 177: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/177.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 178: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/178.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 179: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/179.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 180: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/180.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 181: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/181.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 182: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/182.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 183: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/183.jpg)
33. Antisymmetry
Suppose that k > 2. Then p0, p1, p2 are all distinct. By thedefinition of K , there exists w ∈ Σ∗ such that(p0, p1) .w = (p1, p2), that is, p0 .w = p1, p0 .w
2 = p1 .w = p2.Since A is aperiodic, there exists m such thatp0 .w
m+1 = p0 .wm; we choose the least m with this property.
Observe that m > 1 since p0 .w2 6= p0 .w .
Now we apply wm−1 to each state in the sequence p0, p1, . . . , pk .Since K ∪ D is stable, we get that for all i = 0, 1, . . . , k − 1 either(pi .w
m−1, pi+1 .wm−1) ∈ K or pi .w
m−1 = pi+1 .wm−1. The
choice of m ensures p0 .wm−1 6= p1 .w
m−1 = p0 .wm whence the
new sequence still contains an obstacle to antisymmetry. On theother hand, p1 .w
m−1 = p0 .wm = p0 .w
m+1 = p2 .wm−1, and
therefore, if we retain in the new sequence only the first state fromeach group of adjacent equal states, the sequence becomes shorter.This contradicts the minimality of p0, p1, . . . , pk .
Mikhail Volkov Synchronizing Finite Automata
![Page 184: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/184.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 185: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/185.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 186: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/186.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 187: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/187.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 188: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/188.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 189: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/189.jpg)
34. Using the Order
Thus, �K is a non-trivial partial order. How does it help?We denote by πK the symmetric closure of �K . Clearly, πK is acongruence of the automaton A .The quotient automaton A /πK has < n states and so it has ashort reset word v by the induction assumption (we induct on thenumber of states). In A , the word v sends the whole state set intoa single πK -class, say, T . In order to quickly synchronize T choosea minimum state q ∈ T and a maximum state p ∈ T w.r.t. �K
(p 6= q whenever |T | > 1). Since A is strongly connected, there isa word u1 of length < n such that p . u1 = q. Then for each r ∈ T
such that p �K r we have q = p . u1 �K r . u1 whence q = r . u1 asq is minimal. If still |T . u1| > 1, we can take a maximum statep′ ∈ T . u1 and repeat the process by taking a word u2 of length< n such that p′ . u2 = q, and so on.
Mikhail Volkov Synchronizing Finite Automata
![Page 190: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/190.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 191: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/191.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 192: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/192.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 193: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/193.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 194: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/194.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 195: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/195.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 196: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/196.jpg)
35. Estimations
How long can be a reset word constructed this way?From the minimum-maximum symmetry it follows that the numberof steps is at most |T |
2 . In the case when T = Q (actually, this isthe worst case) we get at most n
2 steps and a word of length atmost n − 1 is added at each step. The resulting reset word is oflength at most n(n−1)
2 . If |T | = m < n, then the quotientautomaton A /πK has at most n −m + 1 states and we first need
a word v of length at most (n−m+1)(n−m)2 to send Q to T and then
a word of length at most m(n−1)2 to synchronize T . It remains to
calculate that
(n −m + 1)(n −m)
2+
m(n − 1)
2≤
n(n − 1)
2
for all m = 2, . . . , n − 1. If m = 1, then v itself is a reset wordfor A .
Mikhail Volkov Synchronizing Finite Automata
![Page 197: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/197.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 198: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/198.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 199: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/199.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 200: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/200.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 201: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/201.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 202: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/202.jpg)
36. Weakly Monotonic Automata
In my paper (Synchronizing automata preserving a chain of partialorders, Theor. Comput. Sci. 410 (2009) 3513–3519) Trahtman’stheorem has been extended to a larger class automata.A DFA A is weakly monotonic of level ℓ if it has a strictlyincreasing chain of stable binary relations ρ0 ⊂ ρ1 ⊂ · · · ⊂ ρℓsatisfying the following conditions:• ρ0 is the equality relation;• for each i = 1, . . . , ℓ, the congruence πi−1 generated by ρi−1 iscontained in ρi and the relation ρi/πi−1 is a partial order onQ/πi−1;• πℓ is the universal relation.
This differs from the notion of a generalized monotonic automatonby just dropping the restriction that the order ρi/πi−1 is linear oneach πi/πi−1-class.
Mikhail Volkov Synchronizing Finite Automata
![Page 203: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/203.jpg)
37. Examples
• every aperiodic automaton is weakly monotonic;
• every automaton with 0 is weakly monotonic (of level 1).
Mikhail Volkov Synchronizing Finite Automata
![Page 204: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/204.jpg)
37. Examples
• every aperiodic automaton is weakly monotonic;
• every automaton with 0 is weakly monotonic (of level 1).
Mikhail Volkov Synchronizing Finite Automata
![Page 205: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/205.jpg)
37. Examples
• every aperiodic automaton is weakly monotonic;
• every automaton with 0 is weakly monotonic (of level 1).
Σ
Mikhail Volkov Synchronizing Finite Automata
![Page 206: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/206.jpg)
38. WM + SC Automata
• Every weakly monotonic automaton with a strongly connectedunderlying digraph is synchronizing. (A non-trivial generalization ofthe corresponding result for aperiodic automata.)
• Every weakly monotonic automaton with a strongly connectedunderlying digraph and n states has a reset word of length
≤⌊
n(n+1)6
⌋
. (This upper bound is new even for the aperiodic case
– recall that Trahtman’s bound was 3 times higher, namely,n(n−1)
2 .)
• Every weakly monotonic synchronizing automaton with n stateshas a reset word of length n(n−1)
2 .
Mikhail Volkov Synchronizing Finite Automata
![Page 207: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/207.jpg)
38. WM + SC Automata
• Every weakly monotonic automaton with a strongly connectedunderlying digraph is synchronizing. (A non-trivial generalization ofthe corresponding result for aperiodic automata.)
• Every weakly monotonic automaton with a strongly connectedunderlying digraph and n states has a reset word of length
≤⌊
n(n+1)6
⌋
. (This upper bound is new even for the aperiodic case
– recall that Trahtman’s bound was 3 times higher, namely,n(n−1)
2 .)
• Every weakly monotonic synchronizing automaton with n stateshas a reset word of length n(n−1)
2 .
Mikhail Volkov Synchronizing Finite Automata
![Page 208: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/208.jpg)
38. WM + SC Automata
• Every weakly monotonic automaton with a strongly connectedunderlying digraph is synchronizing. (A non-trivial generalization ofthe corresponding result for aperiodic automata.)
• Every weakly monotonic automaton with a strongly connectedunderlying digraph and n states has a reset word of length
≤⌊
n(n+1)6
⌋
. (This upper bound is new even for the aperiodic case
– recall that Trahtman’s bound was 3 times higher, namely,n(n−1)
2 .)
• Every weakly monotonic synchronizing automaton with n stateshas a reset word of length n(n−1)
2 .
Mikhail Volkov Synchronizing Finite Automata
![Page 209: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/209.jpg)
38. WM + SC Automata
• Every weakly monotonic automaton with a strongly connectedunderlying digraph is synchronizing. (A non-trivial generalization ofthe corresponding result for aperiodic automata.)
• Every weakly monotonic automaton with a strongly connectedunderlying digraph and n states has a reset word of length
≤⌊
n(n+1)6
⌋
. (This upper bound is new even for the aperiodic case
– recall that Trahtman’s bound was 3 times higher, namely,n(n−1)
2 .)
• Every weakly monotonic synchronizing automaton with n stateshas a reset word of length n(n−1)
2 .
Mikhail Volkov Synchronizing Finite Automata
![Page 210: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/210.jpg)
38. WM + SC Automata
• Every weakly monotonic automaton with a strongly connectedunderlying digraph is synchronizing. (A non-trivial generalization ofthe corresponding result for aperiodic automata.)
• Every weakly monotonic automaton with a strongly connectedunderlying digraph and n states has a reset word of length
≤⌊
n(n+1)6
⌋
. (This upper bound is new even for the aperiodic case
– recall that Trahtman’s bound was 3 times higher, namely,n(n−1)
2 .)
• Every weakly monotonic synchronizing automaton with n stateshas a reset word of length n(n−1)
2 .
Mikhail Volkov Synchronizing Finite Automata
![Page 211: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/211.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
Mikhail Volkov Synchronizing Finite Automata
![Page 212: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/212.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
Mikhail Volkov Synchronizing Finite Automata
![Page 213: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/213.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
Mikhail Volkov Synchronizing Finite Automata
![Page 214: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/214.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
Mikhail Volkov Synchronizing Finite Automata
![Page 215: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/215.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
n − 1 ≤ CSCA(n)
Mikhail Volkov Synchronizing Finite Automata
![Page 216: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/216.jpg)
39. Lower Bounds
A bad news is that there are no matching lower bounds for theupper bounds just discussed. No aperiodic (even no weaklymonotonic) strongly connected automaton is known for which theminimum length of reset words would be greater than or equal tothe number of states.
Let CSCA(n) denote the restriction of the Cerny function to theclass of all strongly connected aperiodic automata, that is,CSCA(n) is the maximum length of shortest reset words for stronglyconnected aperiodic automata with n states. Then our currentknowledge can be summarized as follows:
n− 1 ≤ CSCA(n) ≤
⌊
n(n + 1)
6
⌋
(Volkov, 2009).
Mikhail Volkov Synchronizing Finite Automata
![Page 217: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/217.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
Mikhail Volkov Synchronizing Finite Automata
![Page 218: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/218.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
(Ananichev, 2010) n +⌊n
2
⌋
− 2 ≤ CA(n)
Mikhail Volkov Synchronizing Finite Automata
![Page 219: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/219.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
(Ananichev, 2010) n+⌊n
2
⌋
−2 ≤ CA(n) ≤n(n − 1)
2(Trahtman, 2007).
Mikhail Volkov Synchronizing Finite Automata
![Page 220: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/220.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
(Ananichev, 2010) n+⌊n
2
⌋
−2 ≤ CA(n) ≤n(n − 1)
2(Trahtman, 2007).
1 2 3 4 5 6
0
a
a
aaa
b b b b bb
a
a, b
Mikhail Volkov Synchronizing Finite Automata
![Page 221: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/221.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
(Ananichev, 2010) n+⌊n
2
⌋
−2 ≤ CA(n) ≤n(n − 1)
2(Trahtman, 2007).
1 2 3 4 5 6
0
a
a
aaa
b b b b bb
a
a, b
This is the first automaton in Ananichev’s series that yieldsthe best up to now lower bound for CA(n).
Mikhail Volkov Synchronizing Finite Automata
![Page 222: Synchronizing Finite Automata - Lecture II. Automata with ...math.sjtu.edu.cn/conference/Bannai/2017/data/20170505B/slides.pdf · Lecture II. Automata with Zero and Aperiodic Automata](https://reader033.vdocuments.site/reader033/viewer/2022060314/5f0b80b47e708231d430d67d/html5/thumbnails/222.jpg)
40. Ananichev’s Series
Similarly, if CA(n) denotes the restriction of the Cerny functionto the class of all aperiodic automata, we have:
(Ananichev, 2010) n+⌊n
2
⌋
−2 ≤ CA(n) ≤n(n − 1)
2(Trahtman, 2007).
1 2 3 4 5 6
0
a
a
aaa
b b b b bb
a
a, b
This is the first automaton in Ananichev’s series that yieldsthe best up to now lower bound for CA(n). It has 7 statesand its shortest reset word is a4b3a of length 7 +
⌊
72
⌋
− 2 = 8.
Mikhail Volkov Synchronizing Finite Automata