1 Robust PCPs of Proximity (Shorter PCPs, applications to Coding) Eli Ben-Sasson (Radcliffe) Oded Goldreich (Weizmann & Radcliffe) Prahladh Harsha (MIT)

Slides:



Advertisements
Similar presentations
Low-End Uniform Hardness vs. Randomness Tradeoffs for Arthur-Merlin Games. Ronen Shaltiel, University of Haifa Chris Umans, Caltech.
Advertisements

An Introduction to Randomness Extractors Ronen Shaltiel University of Haifa Daddy, how do computers get random bits?
Lower Bounds for Non-Black-Box Zero Knowledge Boaz Barak (IAS*) Yehuda Lindell (IBM) Salil Vadhan (Harvard) *Work done while in Weizmann Institute. Short.
Russell Impagliazzo ( IAS & UCSD ) Ragesh Jaiswal ( Columbia U. ) Valentine Kabanets ( IAS & SFU ) Avi Wigderson ( IAS ) ( based on [IJKW08, IKW09] )
Direct Product : Decoding & Testing, with Applications Russell Impagliazzo (IAS & UCSD) Ragesh Jaiswal (Columbia) Valentine Kabanets (SFU) Avi Wigderson.
A threshold of ln(n) for approximating set cover By Uriel Feige Lecturer: Ariel Procaccia.
Approximate List- Decoding and Hardness Amplification Valentine Kabanets (SFU) joint work with Russell Impagliazzo and Ragesh Jaiswal (UCSD)
Gillat Kol joint work with Ran Raz Locally Testable Codes Analogues to the Unique Games Conjecture Do Not Exist.
Foundations of Cryptography Lecture 10 Lecturer: Moni Naor.
Locally Decodable Codes from Nice Subsets of Finite Fields and Prime Factors of Mersenne Numbers Kiran Kedlaya Sergey Yekhanin MIT Microsoft Research.
List decoding Reed-Muller codes up to minimal distance: Structure and pseudo- randomness in coding theory Abhishek Bhowmick (UT Austin) Shachar Lovett.
Bounds on Code Length Theorem: Let l ∗ 1, l ∗ 2,..., l ∗ m be optimal codeword lengths for a source distribution p and a D-ary alphabet, and let L ∗ be.
Dana Moshkovitz. Back to NP L  NP iff members have short, efficiently checkable, certificates of membership. Is  satisfiable?  x 1 = truex 11 = true.
Two Query PCP with Subconstant Error Dana Moshkovitz Princeton University and The Institute for Advanced Study Ran Raz The Weizmann Institute 1.
Probabilistically Checkable Proofs (and inapproximability) Irit Dinur, Weizmann open day, May 1 st 2009.
Probabilistically Checkable Proofs Madhu Sudan MIT CSAIL 09/23/20091Probabilistic Checking of Proofs TexPoint fonts used in EMF. Read the TexPoint manual.
Umans Complexity Theory Lectures Lecture 15: Approximation Algorithms and Probabilistically Checkable Proofs (PCPs)
Quantum locally-testable codes Dorit Aharonov Lior Eldar Hebrew University in Jerusalem.
Gillat Kol joint work with Ran Raz Locally Testable Codes Analogues to the Unique Games Conjecture Do Not Exist.
Inapproximability from different hardness assumptions Prahladh Harsha TIFR 2011 School on Approximability.
Two Query PCP with Sub-constant Error Dana Moshkovitz Princeton University Ran Raz Weizmann Institute 1.
Introductions for the “Weizmann Distinguished Lectures Day” by Oded Goldreich.
PCPs and Inapproximability Introduction. My T. Thai 2 Why Approximation Algorithms  Problems that we cannot find an optimal solution.
Complexity 15-1 Complexity Andrei Bulatov Hierarchy Theorem.
Short PCPs verifiable in Polylogarithmic Time Eli Ben-Sasson, TTI Chicago & Technion Oded Goldreich, Weizmann Prahladh Harsha, Microsoft Research Madhu.
Correcting Errors Beyond the Guruswami-Sudan Radius Farzad Parvaresh & Alexander Vardy Presented by Efrat Bank.
6/20/2015List Decoding Of RS Codes 1 Barak Pinhas ECC Seminar Tel-Aviv University.
EXPANDER GRAPHS Properties & Applications. Things to cover ! Definitions Properties Combinatorial, Spectral properties Constructions “Explicit” constructions.
CS151 Complexity Theory Lecture 10 April 29, 2004.
CS151 Complexity Theory Lecture 16 May 25, CS151 Lecture 162 Outline approximation algorithms Probabilistically Checkable Proofs elements of the.
1 Slides by Asaf Shapira & Michael Lewin & Boaz Klartag & Oded Schwartz. Adapted from things beyond us.
CS151 Complexity Theory Lecture 9 April 27, 2004.
Some 3CNF Properties are Hard to Test Eli Ben-Sasson Harvard & MIT Prahladh Harsha MIT Sofya Raskhodnikova MIT.
Dana Moshkovitz, MIT Joint work with Subhash Khot, NYU.
Of 28 Probabilistically Checkable Proofs Madhu Sudan Microsoft Research June 11, 2015TIFR: Probabilistically Checkable Proofs1.
Correlation testing for affine invariant properties on Shachar Lovett Institute for Advanced Study Joint with Hamed Hatami (McGill)
CS151 Complexity Theory Lecture 9 April 27, 2015.
Great Theoretical Ideas in Computer Science.
Sub-Constant Error Low Degree Test of Almost-Linear Size Dana Moshkovitz Weizmann Institute Ran Raz Weizmann Institute.
CS151 Complexity Theory Lecture 13 May 11, Outline proof systems interactive proofs and their power Arthur-Merlin games.
Week 10Complexity of Algorithms1 Hard Computational Problems Some computational problems are hard Despite a numerous attempts we do not know any efficient.
Great Theoretical Ideas in Computer Science.
Communication vs. Computation S Venkatesh Univ. Victoria Presentation by Piotr Indyk (MIT) Kobbi Nissim Microsoft SVC Prahladh Harsha MIT Joe Kilian NEC.
Umans Complexity Theory Lectures Lecture 1a: Problems and Languages.
Probabilistically Checkable Proofs Madhu Sudan MIT CSAIL.
CS151 Complexity Theory Lecture 16 May 20, The outer verifier Theorem: NP  PCP[log n, polylog n] Proof (first steps): –define: Polynomial Constraint.
CS151 Complexity Theory Lecture 15 May 18, Gap producing reductions Main purpose: –r-approximation algorithm for L 2 distinguishes between f(yes)
Umans Complexity Theory Lectures Lecture 16: The PCP Theorem.
Pseudorandomness: New Results and Applications Emanuele Viola IAS April 2007.
RS – Reed Solomon Error correcting code. Error-correcting codes are clever ways of representing data so that one can recover the original information.
Complexity Theory and Explicit Constructions of Ramsey Graphs Rahul Santhanam University of Edinburgh.
1 Tolerant Locally Testable Codes Atri Rudra Qualifying Evaluation Project Presentation Advisor: Venkatesan Guruswami.
Tali Kaufman (Bar-Ilan)
Algebraic Property Testing:
Probabilistic Algorithms
Derandomization & Cryptography
Randomness and Computation
Locality in Coding Theory II: LTCs
Umans Complexity Theory Lectures
Sublinear-Time Error-Correction and Error-Detection
Locality in Coding Theory
Sublinear-Time Error-Correction and Error-Detection
Circuit Lower Bounds A combinatorial approach to P vs NP
Local Error-Detection and Error-correction
Locally Decodable Codes from Lifting
The Curve Merger (Dvir & Widgerson, 2008)
Robust PCPs of Proximity (Shorter PCPs, applications to Coding)
Locality in Coding Theory II: LTCs
Every set in P is strongly testable under a suitable encoding
Umans Complexity Theory Lectures
Presentation transcript:

1 Robust PCPs of Proximity (Shorter PCPs, applications to Coding) Eli Ben-Sasson (Radcliffe) Oded Goldreich (Weizmann & Radcliffe) Prahladh Harsha (MIT) Madhu Sudan (MIT & Radcliffe) Salil Vadhan (Harvard & Radcliffe)

2 NP – Classical Proofs NP – Class of languages that have short proofs of membership NP – Class of languages that have short proofs of membership V (deterministic verifier) Proof x 2 L GraphColoring F ormu l a Á G rap h G Satisfiability Proof = 3 coloring Proof = Satisfying assignment Completeness: Soundness: x 2 L ) 9 ¼ ; V ( x ; ¼ ) = accep t x = 2 L ) 8 ¼ ; V ( x ; ¼ ) = re j ec t ( x _ y _ ¹ z ) ::: ( ¹ x _ y _ z ) ¼

3 PCP Theorem [AS ’92, ALMSS ’92] V (deterministic verifier) V (probabilistic verifier) PCP Theorem NP Proof Completeness: Soundness: x = 2 L ) 8 ¼ ; P r [ V ¼ ( x ) = 1 ] · 1 2 x 2 L ) 9 ¼ ; P r [ V ¼ ( x ) = 1 ] = 1 ¼ Parameters: 1.# random coins - O(log n) 2.# queries - constant 3.proof size - polynomial

4 PCPs - Significance Major impact on the study of combinatorial optimization Major impact on the study of combinatorial optimization Consequence: For many NP-hard combinatorial optimization problems, finding near-optimal solutions is also NP-hard Consequence: For many NP-hard combinatorial optimization problems, finding near-optimal solutions is also NP-hard Approximating MAXSAT to within a factor of (8/7 -  ), for any  > 0, is NP-hard Approximating MAXSAT to within a factor of (8/7 -  ), for any  > 0, is NP-hard (Will not dwell into consequence on combinatorial optimization)

5 Short PCPs? How long is the new PCP proof? How long is the new PCP proof? Old NP proof – n ; New PCP proof - ? Old NP proof – n ; New PCP proof - ? Why Short PCPs? Why Short PCPs? Upper bounds Upper bounds Cryptography Cryptography Computationally Sound Proofs and applications [Kil ’92, Mic ’94, CGH ’98, Bar ’01] Computationally Sound Proofs and applications [Kil ’92, Mic ’94, CGH ’98, Bar ’01] Coding Theory Coding Theory Locally testable codes [GS ’02, BSVW ’03, this paper] Locally testable codes [GS ’02, BSVW ’03, this paper] “Relaxed Locally Decodable Codes” [this paper] “Relaxed Locally Decodable Codes” [this paper]

6 Why Short PCPs? (Contd) Lower Bounds Lower Bounds Tightness of approximation algorithms with respect to running time Tightness of approximation algorithms with respect to running time e.g.: If SAT has a PCP of size n  then e.g.: If SAT has a PCP of size n  then + SAT 2 TIME ¡ 2 ­ ( n ) ¢ Approximating requires time at least MAXSAT 2 n 1 = ®

7 Short PCPs – Earlier Results [PS ’94] [PS ’94] Proof Size = n 1+ , query = O(1/) Proof Size = n 1+ , query = O(1/) (Constant hidden in big-O ¼ 10 6 ) [Hås ’97] [Hås ’97] Proof Size = n , query = 3; Proof Size = n , query = 3;

8 Short PCPs vs Query Complexity queries proof size [HS ’00] [GS ’02, BSVW ’03] This paper This paper n 3 + ² n ¢ 2 p l o g n n ¢ 2 ( l og n ) ² O ( 1 ² ) n ¢ 2 ( l og l og n ) c O ( 1 ) o ( l og l ogn ) 17 ( = n 1 + o ( 1 ) )

9 Our Main Results Main Theorem: Satisfiability of circuits of size n can be probabilistically verified By probing a proof of length By probing a proof of length in bit-locations. in bit-locations.OR By probing a proof of length By probing a proof of length in bit-locations. Previous PCPs required length proof size even when reading bit-locations [GS ’02, BSVW ’03] [GS ’02, BSVW ’03] n ¢ 2 ( l og n ) ² O ( 1 ² ) n ¢ 2 p l o g n n ¢ 2 ( l og l og n ) c o ( l og l ogn ) 2 p l og n

10 Proof Techniques New Definition: Robust PCP of Proximity New Definition: Robust PCP of Proximity New Composition Theorem New Composition Theorem Essential for short PCPs Essential for short PCPs simple, modular simple, modular Building Block Building Block

11 Robust PCP of Proximity and Composition Theorem

12 PCP – Definition (Recall) VLVL (probabilistic verifier) x- T h eorem Completeness: Soundness: x 2 L ) 9 ¼ ; P r [ V ¼ ( x ) = 1 ] = 1 ¼ x = 2 L ) 8 ¼ ; P r [ V ¼ ( x ) = 1 ] · 1 2 Parameters of Interest: size of proof (|   # queries (q ) |  | · q ¢ 2 rand

13 Why Composition? Don’t know to build PCPs with q = O(1) and Don’t know to build PCPs with q = O(1) and size = poly(n) directly size = poly(n) directly However, However, [AS ’92, ALMSS ’92] type of PCP: size = poly(n ) q = poly log n Verifier V [AS ’92, ALMSS ’92] “magically compose” verifier V with itself to obtain new verifier V © V with following parameters size = poly(n ) q = poly log log n V © V

14 Proof Composition, a la [AS ‘92] VLVL r = O ( l ogn ) q = po l y l ogn ¼ x Completeness: Soundness : x 2 L ) 9 ¼ ; P r [ V ¼ ( x ) = 1 ] = 1 x = 2 L ) 8 ¼ ; P r [ V ¼ ( x ) = 1 ] · 1 2 DRDR Consistency Check a 1 a 2 ::: a Q R Need to check if satisfy consistency check D R a 1 a 2 ::: a Q R Idea : Use a PCP verifier to check ! x 2 L ) 9 ¼ ; P r [ D R ( a 1 ::: a Q R ) = 1 ] = 1 x = 2 L ) 8 ¼ ; P r [ D R ( a 1 ::: a Q R ) = 1 ] · 1 2 R an d omco i ns- R

15 Proof Composition, Contd DRDR Consistency Check a 1 a 2 ::: a Q R Create language Check if using a PCP veriifier L R = f ( a 1 ;:::; a Q R ) j D R accep t s ( a 1 ;:::; a Q R ) g ( a 1 ;:::; a Q R ) 2 L R VLVL ¼ x VLRVLR ¼ R Problem: PCP verifier V L R needs to read all of theorem (input) Key Observation: “PCP Verifier barely looks at Theorem” [BFLS ’91] : Assume theorem is encoded and count #queries into theorem

16 [BFLS ’91] PCP Verifier (Holographic Proofs) V (probabilistic verifier) x- T h eorem Completeness: Soundness: ¼ Important: # queries = sum of queries into encoded theorem + proof E nc ( x ) - E nco d i ng x 2 L ) 9 ¼ ; P r [ V E nc ( x ) ; ¼ = 1 ] = 1 y ¡ f ar f rom E nc ( L ) ) 8 ¼ ; P r [ V y ; ¼ = 1 ] · 1 2 y

17 Proof Composition, Contd VLVL x VLRVLR ¼ R a 1 a 2 ::: a Q R ¼ E nc ( a 1 a 2 ::: a Q R ) Problem: Need to check and are consistent. Semantics of arranging this is complex. Earlier performed by “parallelization” – costly in randomness (large proof size) ( a 1 ;:::; a Q R ) E nc ( a 1 a 2 ::: a Q R ) Idea: Remove restriction that theorem is encoded !

18 PCP of Proximity (PCPP) V (probabilistic verifier) x- T h eorem Completeness: Soundness: ¼ # queries = sum of queries into theorem + proof Theorem in un-encoded format  – proximity parameter x 2 L ) 9 ¼ ; P r [ V x ; ¼ = 1 ] = 1 ¢ ( x ; L ) > ± ) 8 ¼ ; P r [ V x ; ¼ () = 1 ] · 1 2 x = 2 L ) 8 ¼ ; P r [ V x ; ¼ () = 1 ] · 1 2

19 Composition again VLVL VLRVLR ¼ R a 1 a 2 ::: a Q R ¼ x Completeness: Soundness : Problem: Need to distinguish between & PCPP distinguishes between & ( a 1 ;:::; a Q R ) 2 L R ( a 1 ;:::; a Q R )= 2 L R ( a 1 ;:::; a Q R ) 2 L R ± ¡ f ar f rom L R Strengthen soundness condition of verifier V L x 2 L ) 9 ¼ ; P r [( a 1 ;:::; a Q R ) 2 L R ] = 1 x = 2 L ) 8 ¼ ; P r [( a 1 ;:::; a Q R )= 2 L R ] > 1 2

20 PCP of Proximity PCP of Proximity V  Completeness: Soundness: DRDR Consistency Check Robust Soundness:  - robustness parameter of robust-PCPP (Robust-PCPP) New! Robust  x a 1 a 2 : a Q R x 2 L ) 9 ¼ ; P r [ D R ( a 1 ;:::; a Q R ) = 1 ] = 1 ¢ ( x ; L ) > ± ) 8 ¼ ; P r [ D R ( a 1 ;:::; a Q R ) = 1 ] · 1 2 ¢ ( x ; L ) > ± ) 8 ¼ ; P r [( a 1 ;:::; a Q R ) i s½- f ar f rom L R ] > 1 2

21 Composition Theorem V OUT  V IN  R1  Rm New PCPP Proof for V COMP = ( ,  R1,…..,  Rm ) V OUT + V IN = V COMP Randomness: r COMP = r OUT + r IN Robustness:  COMP =  IN Proximity:  COMP =  OUT Queries: q COMP = q IN x V IN Req. of Inner Verifier:  IN (proximity) <  OUT (robustness)

22 Advantages of PCPPs Give shortest known PCPs Give shortest known PCPs Allow natural self-composition Allow natural self-composition Simpler constructions of PCPs (no parallelization) Simpler constructions of PCPs (no parallelization) Coding applications: Coding applications: Simple, highly efficient Locally Testable Codes Simple, highly efficient Locally Testable Codes Simple, highly efficient Relaxed Locally Decodable Codes Simple, highly efficient Relaxed Locally Decodable Codes Any efficient property is locally testable (with a little bit of help) Any efficient property is locally testable (with a little bit of help)

23 PCPPs – Brief History Holographic proofs - PCPPs where assignment x Holographic proofs - PCPPs where assignment x is encoded. [BFLS ’91] is encoded. [BFLS ’91] PCPP - implicit in low-degree tests PCPP - implicit in low-degree tests [RS ’92, ALMSS ’92] [RS ’92, ALMSS ’92] PCPPs - special case of “PCP Spot Checkers” PCPPs - special case of “PCP Spot Checkers” [EKR ’99] [EKR ’99] PCPP – extension of Property Testing PCPP – extension of Property Testing [RS ’92, GGR ’96] [RS ’92, GGR ’96] Assignment Testers of [DR ’03] similar to PCPPs. Assignment Testers of [DR ’03] similar to PCPPs.

24 Building Block

25 Robust PCPPs constructions Most existing PCP constructions can be modified to obtain robust PCPs of Proximity Most existing PCP constructions can be modified to obtain robust PCPs of Proximity However, the parameters of such robust-PCPPs do not satisfy our needs However, the parameters of such robust-PCPPs do not satisfy our needs So, build robust PCPP from scratch So, build robust PCPP from scratch

26 Bird’s eye-view of PCP construction F m F m f f 2 PCP Construction: Sequence of function evaluations, f i : F m ! F Checks performed by verifier Each function f i : F m ! F is a low-degree polynomial Input Consistency: f 1 ¼ input Each f i+1 is obtained consistently from f i e.g.: f i+1 (x) = f i (x) ¢ f i (x+1) final function f r :F m ! F is identically zero i.e., f r ´ 0 How to test if a function is a low-degree polynomial ? Input: Evaluation of function f at each point in F m Need to check if evaluation of f is close to the evaluation of a low-degree polynomial f r F m

27 Low Degree Polynomials Main Tool – Low Degree Polynomial over Finite Fields Main Tool – Low Degree Polynomial over Finite Fields (Reed-Muller Codes) (Reed-Muller Codes) F - finite field, f : F m ! F, m-variate polynomial over F, deg( f ) = maximal degree of monomial in f F - finite field, f : F m ! F, m-variate polynomial over F, deg( f ) = maximal degree of monomial in f l f : F m ! F [Schwartz-Zippel] [Schwartz-Zippel] If f  g have degree < d, then If f  g have degree < d, then Fact: Fact: If deg( f ) < d and l – line, then f restricted to line l is a univariate polynomial of degree < d. If deg( f ) < d and l – line, then f restricted to line l is a univariate polynomial of degree < d. P r [ f ( x ) = g ( x )] · d j F j

28 Low Degree Test (LDT) Robust Soundness of LDT [RS ’92, ALMSS ’92] Robust Soundness of LDT [RS ’92, ALMSS ’92] f :F m ! F is  -far from low degree, then Pr l [ f | l is far from being low-degree ] > () Amount of Randomness Required: Amount of Randomness Required: [RS ’92, ALMSS ’92] 2 points – 2 log | F m | [RS ’92, ALMSS ’92] 2 points – 2 log | F m | [BSVW ’03] derandomized set of lines ¼ log | F m | [BSVW ’03] derandomized set of lines ¼ log | F m | Input: Table of evaluations of f at each point of F m Output: Is f low-degree? Choose a random line l Read f along line l Check that restriction of f along l is a univariate low-degree polynomial l f : F m ! F

29 Robust LDTs via Bundling f 1 f 2 f r Each LDT performed separately Each LDT performed separately Possible to cheat by having just one of f i not low-degree Possible to cheat by having just one of f i not low-degree NOT ROBUST NOT ROBUST Bundle evaluations of diff. polys. together and perform LDTs in parallel (bundling) Bundle evaluations of diff. polys. together and perform LDTs in parallel (bundling) PCPP  on query x returns (f 1 (x),f 2 (x),…, f r (x)) PCPP  on query x returns (f 1 (x),f 2 (x),…, f r (x)) Robust over larger alphabet F r Can use error-correcting code to make robust over binary alphabet. Can use error-correcting code to make robust over binary alphabet. PCPP  on query x returns ECC(f 1 (x),f 2 (x),…, f r (x)) PCPP  on query x returns ECC(f 1 (x),f 2 (x),…, f r (x))

30 Building Block - Robust-PCPP Randomness: Randomness: # Queries : # Queries : Robustness Parameter Robustness Parameter Proximity Parameter Proximity Parameter 1 2 l ogn + O ( l og l ogn ) cons t an t p n ¢ po l y l og n

31 Applications to Coding Locally Testable Codes Relaxed-Locally Decodable Codes

32 Locally Testable Codes Lower Bounds Lower Bounds [BHR ’03] Random LDPC Codes are not LTCs [BHR ’03] Random LDPC Codes are not LTCs LTC Constructions LTC Constructions [GS ’02, BSVW ’03] [GS ’02, BSVW ’03] This paper This paper k ¡ ! k ¢ 2 p l og k k ¡ ! k ¢ 2 ( l og k ) ² T constant # queries w w – codeword: w – codeword: Tester T accepts with probability 1 Tester T accepts with probability 1 w - far from codeword: w - far from codeword: Tester T accepts with low probability Tester T accepts with low probability Tester

33 Locally Decodable Codes Hadamard – locally decodable, but poor rate Hadamard – locally decodable, but poor rate Upper Bound: [BIKR ’02] n · 2 O(k) Upper Bound: [BIKR ’02] n · 2 O(k) Lower Bound: [KT ’00] n ¸ k (1) Lower Bound: [KT ’00] n ¸ k (1) D constant # queries c i th mesg bit? r corruption If less than  n bits corrupted, for all message bits i P r [ D r ( i ) = m i ] ¸ 3 4

34 Relaxed Locally Decodable Codes New! This paper: For every  > 0, there exist relaxed- LDCs with This paper: For every  > 0, there exist relaxed- LDCs with D constant # queries c i th mesg bit? r corruption If less than  n bits corrupted, for “most’’ message bits i P r [ D r ( i ) = m i ] ¸ 3 4 For remaining bits P r [ D r ( i ) = ? ] ¸ 3 4 k ¡ ! k 1 + ²

35 Summary of results Defined: Robust PCP of proximity Defined: Robust PCP of proximity Strengthened definition of standard PCPs Strengthened definition of standard PCPs Composition Theorem Composition Theorem simple, modular simple, modular Simpler constructions of PCPs Simpler constructions of PCPs Coding applications: Coding applications: Locally Testable Codes Locally Testable Codes Relaxed Locally Decodable Codes Relaxed Locally Decodable Codes

36 The End