Download presentation

Presentation is loading. Please wait.

Published byReilly Ore Modified over 3 years ago

1
Direct-Product testing Parallel Repetitions And Foams Avi Wigderson IAS

2
Parallel Repetition of Games and Periodic Foams

3
Isoperimetric problem: Minimize surface area given volume. One bubble. Best solution: Sphere

4
Many bubbles Isoperimetric problem: Minimize surface area given volume. Why? Physics, Chemistry, Engineering, Math… Best solution?: Consider R 3 Lord Kelvin 1873 Optimal… Wearie-Phelan 1994 Even better

5
Our Problem Minimum surface area of body tiling R d with period Z d ? Volume=1 d=2 area: 4>4Choe89: Optimal!

6
Bounds in d dimensions OPT [Kindler,ODonnell, Rao,Wigderson] Rao,Wigderson] OPT Spherical Cubes exist! Probabilistic construction! (simpler analysis [Alon-Klartag]) OPEN: Explicit?

7
Randomized Rounding Round points in R d to points in Z d such that for every x,y 1. 2. x y 1 Bound does not depend on d

8
Spine Torus Surface blocking all cycles that wrap around

9
Probabilistic construction of spine Step 1 Randomly construct B in [0,1) d, which in expectation satisfies B Step 2 Sample independent translations of B until [0,1) d is covered, adding new boundaries to spine.

10
PCPs & Linear equations over GF(2) m linear equations: Az = b in n variables: z 1,z 2,…,z n Given (A,b) 1) Does there exist z satisfying all m equations? Easy – Gaussian elimination 2) Does there exist z satisfying.9m equations? NP-hard 3) Does there exist z satisfying.5m equations? Easy – YES! [Hastad] δ >0, it is NP-hard to distinguish (A,b) which are not (½+ δ )-satisfiable, from those (1- δ )-satisfiable!

11
Linear equations as Games 2n variables: X 1,X 2,…,X n, Y 1,Y 2,…,Y n m linear equations: Xi 1 + Yi 1 = b 1 Xi 2 + Yi 2 = b 2 ….. Xi m + Yi m = b m Promise: no setting of the X i,Y i satisfy more than (1- δ )m of all equations Game G Draw j [m] at random Xi j Yi j Alice Bob α j β j Check if α j + β j = b j Pr [YES] 1- δ

12
Hardness amplification by parallel repetition 2n variables: X 1,X 2,…,X n, Y 1,Y 2,…,Y n m linear equations: Xi 1 + Yi 1 = b 1 Xi 2 + Yi 2 = b 2 ….. Xi m + Yi m = b m Promise: no setting of the X i,Y i satisfy more than (1- δ )m of all equations Game G k Draw j 1,j 2,…j k [m] at random Xi j1 Xi j2 Xi jk Yi j1 Yi j2 Yi jk Alice Bob α j1 α j2 α jk β j1 β j2 β jk Check if α jt + β jt = b jt t [k] Pr[YES] (1- δ 2 ) k [Raz,Holenstein,Rao] Pr[YES] (1- δ 2 ) k [Feige-Kindler-ODonnell] Spherical Cubes [Raz] X [KORW]Spherical Cubes

13
Hardness amplification by other means? Xi 1 + Yi 1 = b 1 Xi 2 + Yi 2 = b 2 ….. Xi m + Yi m = b m Promise: no setting of the X i,Y i satisfy more than (1- δ )m of all equations Amplification Xi j1 … Xi jk Yi j1 … Yi jk Alice Bob α j1 … α jk β j1 … β jk Test: α jt + β jt = b jt t ? Pr[YES] (1- δ 2 ) k [Raz,Holenstein,Rao] Pr[YES] (1- δ 2 ) k [Raz] Major open question: Is there Test s.t. Pr[YES] (1- δ ) k ? [Khot] Unique games conjecture Idea: force each player to answer consistently - e.g. make Alice commit to one assignment of Xis [ Impagliazzo-Kabanets-W ] New Test with Pr[YES] (1- δ )k

14
Direct-product testing Part of - local testing of codes - property testing - discrete rigidity / stability Related to - local decoding of codes - Yaos XOR lemma

15
Direct Product: Definition For f : U R, the k -wise direct product f k : U k R k is f k (x 1,…, x k ) = ( f(x 1 ), …, f(x k ) ). [Impagliazzo02, Trevisan03]: DP Code TT ( f k ) is DP Encoding of TT ( f ) Rate and distance of DP Code are bad, but the code is still very useful in Complexity …

16
Direct-Product Testing Given an oracle C : U k R k Test makes few queries to C, and (1) Accept if C = f k. (2) Reject if C is far away from any f k (2) [Inverse Thm] Pr [ Test accepts C ] > C f k on > ( ) of inputs. - Minimize # queries e.g. 2, 3,.. ? - Analyze small e.g. < 1/k, < exp(-k) ? - Reduce rate/Derandomize e.g.|C| = poly (|U|) ?

17
DP Testing History Given an oracle C : U k R k, is C ¼ g k ? #queries acc prob Goldreich-Safra00* 20.99 Dinur-Reingold06 2.99 Dinur-Goldenberg08 2 1/k α Dinur-Goldenberg08 2 1/k Impagliazzo-Kabanets-W08 3 exp(-k α ) Impagliazzo-Kabanets-W08* 2 1/k α / *Derandomization

18
Consistency tests

19
V-Test [GS00,FK00,DR06,DG08] Pick two random k-sets S 1 = (B 1,A), S 2 = (A,B 2 ) with m = k 1/2 common elements A. Check if C(S 1 ) A = C(S 2 ) A B1B1 B2B2 A [DG08]: If V-Test accepts with probability ² > 1/k α, then there is g : U R s.t. C ¼ g k on at least ² / 4 fraction of k-sets. [IKW09]: Derandomize [DG08]: V-Test fails for ² <1/k S1S1 S2S2

20
Z-Test Pick three random k-sets S 1 =(B 1, A 1 ), S 2 =(A 1,B 2 ), S 3 =(B 2, A 2 ) with |A 1 | = |A 2 | = m = k 1/2. Check if C(S 1 ) A 1 = C(S 2 ) A 1 and C(S 2 ) B 2 = C(S 3 ) B 2 Theorem [IKW09]: If Z-Test accepts with probability ² > exp(-k α ), then there is g : U R s.t. C ¼ g k on at least ² / 4 fraction of k-sets. B1B1 B2B2 A1A1 A2A2 S1S1 S2S2 S3S3

21
Proof Ideas

22
Proof steps 1. Pr [ Test accepts C ] > structure 2. Structure local agreement 3. local agreement global agreement Agreement: there is g : U R s.t. C ¼ g k on at least ² / 4 fraction of k-sets.

23
Flowers, cores, petals Flower: determined by S=(A,B) Core: A Core values: α =C(A,B) A Petals: Cons A, B = { (A,B) | C(A,B) A = α } In a flower, all petals agree on core values! [IJKW08]: Flower analysis for DP-decoding. Symmetry arguments! B B4B4 AA B2B2 B3B3 B1B1 B5B5

24
V-Test ) Structure (similar to [FK, DG]) Suppose V-Test accepts with probability ². Cons A, B = { (A,B) | C(A,B) A = C(A,B) A } (1) Largeness: Many ( ² /2) flowers (A,B) have many ( ² /2) petals Cons A, B (2) Harmony: In every large flower, almost all pairs of overlapping sets in Cons are almost perfectly consistent. B B4B4 AA B2B2 B3B3 B1B1 B5B5

25
V-Test: Harmony Almost all B 1 = (E,D 1 ) and B 2 = (E,D 2 ) in Cons (with |E|=|A|) satisfy C(A, B 1 ) E C(A, B 2 ) E B D2D2 D1D1 A E Proof: Symmetry between A and E (few errors in AuE ) Chernoff: ² ¼ exp(-k α ) E A Implication: Restricted to Cons, an approx V-Test on E accepts almost surely: Unique Decode!

26
Harmony ) Local DP Main Lemma: Assume (A,B) is harmonious. Define g(x) = Plurality { C(A,B) x | B 2 Cons & x 2 B } Then C(A,B) B ¼ g k (B), for almost all B 2 Cons B AA D2D2 D1D1 E Intuition: g = g (A,B) is the unique (approximate) decoding of C on Cons (A,B) B x Idea: Symmetry arguments. Largness guarantees that random selections are near-uniform. Challenge: Our analysis gets stuck in ² ¼ exp(-k) Can one get ² ¼ exp(-k) ??

27
Local DP structure across U k Field of flowers (A i,B i ) For each, g i s.t C(S) ¼ g i k (S) if S 2 Cons (Ai,Bi) Global g? B2B2 AA2A2 BiBi AAiAi B AA B3B3 AA3A3 B1B1 AA1A1

28
From local DP to global DP Q: How to glue local solutions ? A: If a typical S has two disjoint large, harmonious As ² > 1/k α high probability (2 queries) [DG] ² > exp(-k α ) Z-test (3 queries) [IKW]

29
Derandomization DP code whose length is poly (|U|), instead of |U| k

30
Inclusion graphs are Samplers Most lemmas analyze sampling properties m-subsets A Subsets: Chernoff bounds – exponential error Subspaces: Chebychev bounds – polynomial error Cons S k-subsets x elements of U

31
Derandomized DP Test Derandomized DP: U=(F q ) d Encode f k (S), S subspace of const dimension (as [IJKW08] ) Theorem (Derandomized V-Test): If derandomized V-Test accepts C with probability ² > poly(1/k), then there is a function g : U R such that C (S) ¼ g k (S) on poly( ² ) of subspaces S. Corollary: Polynomial rate testable DP-code with [DG] parameters!

32
Summary Spherical cubes exist Power of consistency

33
Counterexample [DG] For every x 2 U pick a random g x : U R For every k-subset S pick a random x(S) 2 S Define C(S) = g x(S) (S) C(S 1 ) A =C(S 2 ) A iff x(S 1 )=x(S 2 ) V-test passes with high prob: ² = Pr[C(S 1 ) A =C(S 2 ) A ] ~ m/k 2 No global g if ² < 1/k 2 B1B1 B2B2 A S1S1 S2S2

Similar presentations

OK

Umans Complexity Theory Lectures Lecture 15: Approximation Algorithms and Probabilistically Checkable Proofs (PCPs)

Umans Complexity Theory Lectures Lecture 15: Approximation Algorithms and Probabilistically Checkable Proofs (PCPs)

© 2018 SlidePlayer.com Inc.

All rights reserved.

Ads by Google

Download ppt on teamviewer 6 Ppt on personality development for school students One act play ppt on website Ppt on bluetooth architecture piconets Ppt on credit default swaps pricing Ppt on idiopathic thrombocytopenic purpura definition Ppt on accounting standard 13 Ppt on asp dot net Free download ppt on data handling in maths for class 8 Ppt on total internal reflection critical angle