constraint satisfaction problem (answer/do something) 4 10/30/2017 3 agents with knowledge and...
TRANSCRIPT
10/30/2017
1
Logical AgentsDeepak Kumar
October-November 2017
What is AI? – Recall from Week 1 & 2
Reasoning with Logic
• Aristotle: What are correct arguments/thought processes?
• Formal Logics:
Socrates is human.All humans are mortal.Therefore Socrates is mortal.
• Laws of thought govern the operation of the mind.
Machines with logic“laws of thought”
Logic
Reasoning Etc.
Thinking Rationally
2
10/30/2017
2
Computer Science - Databases
• Database Systems
Database
Data
Query
Response
3
Agents with Knowledge and Reasoning
• Knowledge-Based Systems
KnowledgeBase
“Knowledge”
Tell/Sense/Ask
Response (Answer/Do Something)
4
10/30/2017
3
Agents with Knowledge and Reasoning
• Knowledge-Based Systems
• Knowledge – set of sentences that describe facts about the world (or domain)
KnowledgeBase
“Knowledge”
Tell/Sense/Ask
Response (Answer/Do Something)
5
Agents with Knowledge and Reasoning
• Knowledge Representation & Reasoning (KRR) Systems
• Knowledge – set of sentences that describe facts about the world (or domain)
• Inferences – procedures/rules that operate on facts to infer new facts
KnowledgeBase
“Knowledge”+
Inference Engine
Tell/Sense/Ask
Response (Answer/Do Something)
6
10/30/2017
4
Examples – Rule-Based Systems
• MYCIN
if stain of organism is gram positive, andmorphology of organism is coccus, andthe growth confirmation of organism is chains
thenid of organism is streptococcus
7
Examples – KRR Systems
• BLOCKSWORLD with declarative knowledge
if a block is on top of another blockthen
the latter block is not clear
Facts
blue block is on top of red blockred block is on the tableblue block is clear
∴ red block is not clear
8
10/30/2017
5
Knowledge Representation
• Syntax: How sentences are formed
• Semantics: Meaning of sentences
• Computation: How sentences are manipulated
Facts
blue block is on top of red blockred block is on the tableblue block is clear
∴ red block is not clear
9
Knowledge Representation
red block is not clear
KR Sentences KR Sentences
Facts about the world Fact about the world
inference
follows
World
Representation
sem
anti
cs
sem
anti
cs
blue block is on top of red blockred block is on the tableblue block is clear
10
10/30/2017
6
Knowledge Representation
red block is not clear
KR Sentences KR Sentences
Facts about the world Fact about the world
inference
follows
World
Representation
sem
anti
cs
sem
anti
cs
blue block is on top of red blockred block is on the tableblue block is clear
Logic is one way of doing this!11
What is a logic?
• Study of correct inferences
Premises
ቋ
………
True
Conclusion
∴ ሽ… is also True
If it is cold then my car will not start.My car will not start.
∴ It is cold.
If it is cold then my car will not start.It is cold.
∴ My car will not start.
12
10/30/2017
7
What is a logic?
• Study of correct inferences – Truth preserving consequences
Premises
ቋ
………
True
Conclusion
∴ ሽ… is also True
If it is cold then my car will not start.My car will not start.
∴ It is cold.
If it is cold then my car will not start.It is cold.
∴ My car will not start.
13
What is a logic?
• Study of correct inferences – Truth preserving consequences
Premises
ቋ
………
True
Conclusion
∴ ሽ… is also True
If it is cold then my car will not start.My car will not start.
∴ It is cold.
If it is cold then my car will not start.It is cold.
∴ My car will not start.
This is NOTTruth preserving.
This isTruth preserving.
14
10/30/2017
8
What is a logic?
• Study of correct inferences – Truth preserving consequences
Premises
ቋ
………
True
Conclusion
∴ ሽ… is also True
If it is cold then my car will not start.My car will not start.
∴ It is cold.
If it is cold then my car will not start.It is cold.
∴ My car will not start.
This is NOTTruth preserving.This is sometimesused in diagnosis!
This isTruth preserving.
15
What is a logic?
• Study of correct inferences
• Formalize the notion of correct inference
• Step1: Define a formal language to write sentences – syntaxwell-formed sentences (wffs)
• Step 2: What do the wffs mean? – Semantics/Model TheoryNeed an interpretation for wffs
• Step 3: Rules of Inference – Proof Theory
16
10/30/2017
9
Rules of Inference red block is not clearKR Sentences KR Sentences
Facts about the world Fact about the world
inference
follows
World
Representation
sem
anti
cs
sem
anti
cs
blue block is on top of red blockred block is on the tableblue block is clear
Properties
Soundness: Every new sentence thatcan be derived from KR is avalid consequence.
Completeness: Every valid consequence ofKR can be formally derived.
17
Rules of Inference red block is not clearKR Sentences KR Sentences
Facts about the world Fact about the world
inference
follows
World
Representationse
man
tics
sem
anti
cs
blue block is on top of red blockred block is on the tableblue block is clear
Properties
Soundness: Every new sentence thatcan be derived from KR is avalid consequence.
Completeness: Every valid consequence ofKR can be formally derived.
This is easy to show
This is true for some logics.
18
10/30/2017
10
There are many types of logic
• Propositional Logic
• First-Order Logic
• Second-Oder Logic
• Temporal Logic
• Modal Logic
• Constraint Logic
• Etc.
19
There are many types of logic
• Propositional Logic
• First-Order Logic
• Second-Oder Logic
• Temporal Logic
• Modal Logic
• Constraint Logic
• Etc.
20
10/30/2017
11
Propositional Logic
• A language for symbolic reasoning
• Proposition – a statement that is either True or False.
E.g.
Bryn Mawr College is located in Canada.Today is a sunny day.23 is a prime number.This class is boring.
21
Propositional Logic - Syntax• Symbols
• Atoms:• Constants - 𝑇𝑟𝑢𝑒 (𝑇), 𝐹𝑎𝑙𝑠𝑒 (𝐹)• Propositional Symbols – 𝑃, 𝑄, 𝑅, 𝑅𝐴𝐼𝑁𝐼𝑁𝐺, 𝐶𝐿𝑂𝑈𝐷𝑌, 𝐵𝑂𝑅𝐼𝑁𝐺, etc.
• Connective Symbols - ∨, ∧, ¬, ⇒
• Sentences – Well-formed formulas (wffs)• Any atom is a wff [Atomic Sentences]
e.g. 𝑃, 𝑄, 𝑅, 𝑅3• Complex Sentences
• If ω1and ω2 are wffs, then so areω1ω1∨ ω2 disjunctionω1∧ ω2 conjunctionω1⇒ ω2 implication¬ ω1 negation
• There are no other wffs.22
10/30/2017
12
Propositional Logic - Syntax• Sentences – Well-formed formulas (wffs)
• Any atom is a wff [Atomic Sentences]e.g. 𝑃, 𝑄, 𝑅, 𝑅3
• Complex Sentences• If ω1and ω2 are wffs, then so are
ω1ω1∨ ω2 disjunctionω1∧ ω2 conjunctionω1⇒ ω2 implication¬ ω1 negation
• There are no other wffs.
23
Examples
𝑃(𝑃 ∨ 𝑄) ⇒ ¬P𝑃 ∧ 𝑄¬ ¬ PP ⇒⇒ ¬P
etc.
Propositional Logic - Syntax• Sentences – Well-formed formulas (wffs)
• Any atom is a wff [Atomic Sentences]e.g. 𝑃, 𝑄, 𝑅, 𝑅3
• Complex Sentences• If ω1and ω2 are wffs, then so are
ω1ω1∨ ω2 disjunctionω1∧ ω2 conjunctionω1⇒ ω2 implication¬ ω1 negation
• There are no other wffs.
24
Examples
𝑃(𝑃 ∨ 𝑄) ⇒ ¬P𝑃 ∧ 𝑄¬ ¬ P
P ⇒⇒ ¬P X
etc.