1. 2 overview of the previous lecture gap-qs[o(n), ,2| | -1 ] gap-qs[o(1), ,2| | -1 ] qs[o(1), ]...
TRANSCRIPT
![Page 1: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/1.jpg)
1
![Page 2: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/2.jpg)
2
Overview of the Previous Overview of the Previous LectureLecture
Gap-QS[O(n),,2||-1]
Gap-QS[O(1),,2||-1]
QS[O(1),]
Solvability[O(1),] 3-SAT
This will imply a strong PCP
characterization of NP
??
![Page 3: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/3.jpg)
3
AimAim
To reduce Gap-QS[O(n),,2||-1] to
Gap-QS[O(1),,2||-1].
p1(x1,...,xn) = 0
p2(x1,...,xn) = 0
...
pn(x1,...,xn) = 0
q1(x16,x21,x32) = 0
q2(x13,x26) = 0
q3(x1,xn,x5n) = 0
...
qm(x3,x4,x5,xn+1) = 0
each equation may depend on many variables
each equation depends only on a constant number of variables
![Page 4: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/4.jpg)
4
New VersionNew Version
Definition (Gap-QS*[D, ,]): Instance: a set of n conjunctions of constant
number of quadratic equations (polynomials) over . Each equation depends on at most D variables.
Problem: to distinguish between:
There is an assignment satisfying all the conjunctions.
No more than an fraction of the conjunctions can be satisfied simultaneously.
![Page 5: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/5.jpg)
5
Conjunctions Rather Than Just Conjunctions Rather Than Just EquationsEquationsAn example for a NO instance of Gap-QS*[1,Z2,½]
x = 0 y = 0
x = 0 y = 1
x = 1 y = 0
Nevertheless, we can satisfy more than a half of the
equations!!
Henceforth, we’ll assume the number of equations in all the
conjunctions is the same. Is this a restriction?
![Page 6: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/6.jpg)
6
RelaxationRelaxation
Claim: Gap-QS*[O(1),,2||-1] reduces to Gap-QS[O(1),,3||-1] (When || is at most polynomial in the size of the input).
Proof: Given an instance of Gap-QS*[O(1),,2||-1], replace each conjunction with all linear combinations of its polynomials.
Make sure that:1) The number of linear combinations over is polynomial.2) The dependency of each linear combination is constant.
![Page 7: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/7.jpg)
7
Correctness of the ReductionCorrectness of the Reduction
If the original system had a common solution, so does the new system.
Otherwise, fix an assignment to the variables of the system and observe the two instances:
![Page 8: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/8.jpg)
8
AnalysisAnalysis
2||-1
fraction of
unsatisfied
conjunctions
fraction of satisfied
conjunctions
polynomials originating
from the blue set
polynomials originating from
the pink set
all satisfied
2||-1
fraction of satisfied polynomials originating
from unsatifiable conjunctions ||-1
![Page 9: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/9.jpg)
9
RelaxationRelaxation
Yes instance of Gap-QS*[O(1),,2||-1] are transformed into Yes instances of Gap-QS[O(1),,3||-1].
No instance of Gap-QS*[O(1),,2||-1] are transformed into No instances of Gap-QS[O(1),,3||-1].
The construction is efficient when || is at most polynomial in the size of the input.
What proves the claim.
![Page 10: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/10.jpg)
10
AmplificationAmplification
Claim: Gap-QS[O(1),,3||-1] reduces to Gap-QS[O(1),,2||-1] (When ||>3 is at most polynomial in the size of the input).
Proof: Given an instance of Gap-QS[O(1),,3||-1], generate the set of all linear combinations of N polynomials.
Make sure that:1) The number of linear combinations over is polynomial.2) The dependency of each linear combination is constant.
A constant to be determined
later.
![Page 11: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/11.jpg)
11
Correctness of the ReductionCorrectness of the Reduction
Again if the original system had a common solution, so does the new system.
Otherwise, fix an assignment and observe a linear combination. – the probability all the N polynomials are
satisfied (3||-1)N
– if not all the polynomials are satisfied, the probability the combination is satisfied ||-1
– When ||>3 for N=4 we get the desired error probability.
![Page 12: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/12.jpg)
12
What Have We Done So Far?What Have We Done So Far?
Gap-QS[O(n),,2||-1]
Gap-QS[O(1),,2||-1]
Gap-QS*[O(1),,2||-1]
Gap-QS[O(1),,3||-1]
QS[O(1),]
Solvability[O(1),] 3-SAT
This will imply a strong PCP
characterization of NP ??
![Page 13: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/13.jpg)
13
New AimNew Aim
To reduce Gap-QS[O(n),,2||-1] to Gap-QS*[O(1),,2||-1].
p1(x1,...,xn) = 0
p2(x1,...,xn) = 0
...
pn(x1,...,xn) = 0
q1(xn,x2n)=0 q2(x1,x2)=0
q3(x13)=0 q2(x12,x234)=0
q4(xn)=0 q5(x1)=0 q4(x1)=0
...
qm(x3,x4,x5,xn+1)=0 qn(x1)=0
each equation may depend on many variables
each conjunction is composed of O(1) equations, which depend only on O(1) variables
![Page 14: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/14.jpg)
14
Rewriting the EquationsRewriting the Equations
Every quadratic polynomial P(x1,...,xD) can be written as
for some series of coefficients {cij}, where
(1) aij=xixj for any 1ijD
(2) ai0=xi for any 1iD
(3) a00=1
Dji ,0
ijij acdepends on the assignment
depends on the polynomial
![Page 15: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/15.jpg)
15
Representing Polynomials As Representing Polynomials As SumsSums For any given point xd we can associate
each aij with a point in Hd (H is some finite field, d=O(log|H|D)).
Now the evaluation of P in this point can be written as
A(h)(h)cdHh
P
![Page 16: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/16.jpg)
16
Representing Polynomials As Representing Polynomials As SumsSumsThe evaluation of every quadratic polynomial at
a point can be written as
where fP(h)=LDEcP(h)·LDEA(h).
dHh
P(h)f
Its total degree is at most 2·d·(|H|-1)
![Page 17: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/17.jpg)
17
Partial SumsPartial Sums
Define:
Hh
d1jj1Hh
d1fd1j
)h,...,h,a,...,f(a...)a,...,a(j,Sum
Sumf can be thought of as a polynomial of total degree at most
2·d2·(|H|-1)
![Page 18: 1. 2 Overview of the Previous Lecture Gap-QS[O(n), ,2| | -1 ] Gap-QS[O(1), ,2| | -1 ] QS[O(1), ] Solvability[O(1), ] 3-SAT This will imply a](https://reader037.vdocuments.site/reader037/viewer/2022110401/56649e215503460f94b0e060/html5/thumbnails/18.jpg)
18
Verifying the Polynomial ZeroesVerifying the Polynomial Zeroes