This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 6.841 Advanced Complexity Theory April 18, 2007 Lecture 18 Lecturer: Madhu Sudan Scribe: Nadia Benbernou 1 Probabilistically Checkable Proofs (PCP) The goal of a probabilistically checkable proof is to verify a proof by looking at only a small number of bits, and probabilistically decide whether to accept or reject. The two resources which PCPs rely on are randomness and queries. A Restricted ( r ( n ) , q ( n ) , a ( n )) PCP verifier is a probabilistic polynomial time verifier with oracle access to a proof π that 1. uses at most r ( n ) random bits. 2. makes q ( n ) queries to π . 3. expects answers of size a ( n ). Definition 1 A language L is in PCP s [ r ( n ) , q ( n ) , a ( n )] if there exists a re stricted ( r ( n ) , q ( n ) , a ( n )) PCP verifier V with soundness parameter s such that: x ∈ L ⇒ ∃ π s.t. Pr[ V π ( x ) accepts ] = 1 x 6∈ L ⇒ ∀ π, Pr[ V π ( x ) accepts ] < s. Theorem 2 (PCP Theorem) There exists a global constant Q such that ∀ L ∈ NP there is a constant c such that L ∈ PCP 1 / 2 [ c log n, Q, 2] The PCP Theorem was first proved by Arora, Safra, Arora, Lund, Motwani, Sudan, and Szegedy, then later by Dinur. It is easy to see that NP = ∪ c ∈ N PCP [0 , n c , 2], since the proof of membership to a language in NP is polynomial in size, so can just query entire proof and then accept or reject. We also have NP = ∪ c ∈ N PCP 1 / 2 [ c log n, n c , 2], to see the forward inclusion that NP ⊆ ∪ c ∈ N PCP 1 / 2 [ c log n, n c , 2] can simulate log n bits of randomness (just enumerate over all random strings, count the number of accepting and rejecting configurations, and then output decision). 2 MAXkSAT Definition 3 The promise problem MAXkSAT is given by: Π Yes = { φ  φ is a kcnf formula and all clauses can be simulaneously satisfied } Π No = { φ  φ is a kcnf formula and any assignment satisfies < 1 fraction of the clauses } 181 Corollary 4 There are constants k, > such that MAX k SAT is NPhard to approximate within 1 ....
View
Full
Document
This note was uploaded on 04/02/2010 for the course CS 6.841 taught by Professor Madhusudan during the Spring '09 term at MIT.
 Spring '09
 MadhuSudan

Click to edit the document details