AI summary by TwinMind

Lecture 4: PCP via GKR and Interactive Arguments, Part 1

MIT OpenCourseWare

1h 29m 42s

Introduction to PCPs

  • Definition: Probabilistically checkable proofs allow verification by reading only few random locations
  • Key idea: Convert classical proof (witness) into larger proof where verifier queries few bits randomly
  • Properties needed: completeness, soundness, efficiency
  • PCP complexity class defined by parameters: C (completeness), S (soundness), R (randomness), Q (queries)

PCP Complexity Class Definition

  • Language L has PCP if probabilistic polynomial-time verifier V exists
  • Completeness: For X ∈ L, ∃ PCP π such that V^π(X) accepts with probability ≥ C
  • Soundness: For X ∉ L, ∀ π, V^π(X) accepts with probability ≤ S
  • Efficiency constraints: V uses R bits randomness, Q oracle calls (non-adaptive)
  • Non-adaptive: verifier chooses all query locations before seeing oracle responses

Main Theorem

  • Theorem: NP ⊆ PCP[1, 1/2, poly log n, poly log n]
  • Can construct PCP with completeness 1, soundness 1/2, poly log randomness/queries
  • Construction focuses on 3SAT (NP-complete language)
  • Circuit verification has poly log depth, log space uniform properties

GKR Protocol Review

  • Circuit has layers with low-degree extensions V_i tilde over field F
  • Parameters: H^m = s (circuit size), F size = poly log s
  • Each layer extended to F^m, maintaining polynomial size
  • Two sum-check protocols per layer, D layers total
  • Sum-check reduces checking layer i to layer i-1

PCP Construction from GKR

  • PCP contains: All V_i tilde for every layer i, every point Z ∈ F^m
  • For every layer i, point Z, randomness R: include all sum-check transcripts
  • Tree structure: F^{3M} possible transcripts per sum-check (3M variables)
  • Size: D × F^m × F^{3M} = polynomial (since F = poly log s)

Verification Process

  • Verifier simulates GKR protocol using oracle
  • Reads output layer, expects value 1
  • Runs 2D sum-checks using oracle responses
  • Checks consistency between layers
  • Requires low-degree test for input layer V_0 tilde

Soundness Analysis

  • Each sum-check has soundness error ≤ 3MH/F
  • Total error: 2D × 3MH/F (union bound over 2D sum-checks)
  • For soundness 1/2: need F > 12DMH
  • Since D, M, H all poly log s, can choose F = poly log s
  • Maintains polynomial PCP size

Complexity Analysis

  • Randomness complexity: 2D sum-checks × 3M log F bits = poly log n
  • Query complexity: 2D × 3M rounds × poly H elements = poly log n
  • PCP size: D × F^m × F^{3M} = polynomial in circuit size
  • Efficient construction: given witness W, can compute PCP in polynomial time

Parameter Engineering

  • Critical choice: H = log n, M = log n / log log n
  • Ensures H^m = n, F^m = poly n
  • Avoids exponential blowup in PCP size
  • Trade-off between field size and number of variables

Low-Degree Testing

  • Must verify V_0 tilde (input layer) is actually low-degree extension
  • In interactive GKR, verifier computes this himself
  • In PCP setting, prover provides it, requires verification
  • Other layers tested implicitly through GKR soundness

Theoretical Significance

  • Known results: can achieve 3-query PCPs with constant soundness
  • This construction gives poly log query complexity
  • Demonstrates power of sum-check protocol as fundamental building block
  • Bridges interactive proofs and non-interactive verification

TwinMind

Get summaries like this for your own meetings

TwinMind takes notes in the background during meetings, lectures, and calls, then writes the summary for you.

Free plan to get started

No bot joins your calls

Phone, desktop & Chrome apps

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome