[Paper Review] A Proof Checking View of Parameterized Complexity

Luke Mathieson|arXiv (Cornell University)|Jun 12, 2012

Complexity and Algorithms in Graphs29 references3 citations

TL;DR

This paper introduces a parameterized proof checking framework by extending the PCP Theorem to parameterized complexity classes such as W[1], W[2], M[1], and the A-hierarchy. It establishes p-PCP characterizations for these classes using probabilistic verifiers that read only polylogarithmic bits of a proof, with query and randomness complexity depending on the parameter k and input size n, thereby enabling new inapproximability results in parameterized complexity.

ABSTRACT

The PCP Theorem is one of the most stunning results in computational complexity theory, a culmination of a series of results regarding proof checking it exposes some deep structure of computational problems. As a surprising side-effect, it also gives strong non-approximability results. In this paper we initiate the study of proof checking within the scope of Parameterized Complexity. In particular we adapt and extend the PCP[n log log n, n log log n] result of Feige et al. to several parameterized classes, and discuss some corollaries.

Motivation & Objective

• To develop a proof-checking perspective for parameterized complexity, analogous to the PCP Theorem in classical complexity.
• To extend the PCP framework to parameterized classes such as W[1], W[2], M[1], and the A-hierarchy, enabling probabilistic verification of solutions with limited access to proofs.
• To establish p-PCP characterizations for these classes with query and randomness complexity depending on the parameter k and input size n.
• To explore the implications of these results for parameterized inapproximability and approximation theory.
• To lay the foundation for future work on tighter p-PCP constructions and extensions to PSPACE-related parameterized classes.

Proposed method

• Adapts the PCP[O(log n), O(1)] framework to parameterized classes by introducing parameterized query and randomness complexity.
• Constructs a probabilistic verifier that checks truth assignments to variable sets in a parameterized formula, using a proof string encoding truth assignments.
• Reduces instances of W[1] and W[2] to weighted satisfiability problems (WSAT) with bounded alternation, then applies PCP-style verification protocols.
• Uses a recursive structure of propositional formulas (Γt,d and Δt,d) to model parameterized classes and applies multilinearity testing and low-degree verification techniques.
• Applies reductions from parameterized problems to WSAT(Γ⁻₁,₂) and AWSAT(Γ⁻₁,₂) to derive p-PCP bounds with parameter-dependent query and randomness complexity.
• Employs a field-based encoding of truth assignments and leverages the structure of even- and odd-indexed variable sets to simulate layered verification.

Experimental results

Research questions

• RQ1 Can the PCP Theorem be meaningfully extended to parameterized complexity classes such as W[1] and W[2]?
• RQ2 What is the minimal query and randomness complexity required for a probabilistic verifier to check proofs for W[1]-complete problems?
• RQ3 How can the PCP framework be adapted to handle the parameterized structure of problems in the A-hierarchy and AW[*]?
• RQ4 Can the resulting p-PCP characterization be used to derive new inapproximability results in parameterized complexity?
• RQ5 What are the limitations of extending this framework to classes with unbounded alternation or higher-order quantification?

Key findings

• The paper establishes that W[1] is contained in p-PCP[O(f(k)n^{g(k)}·log n·log log n), O(f(k)n^{g(k)}·log n·log log n)], where f and g are computable functions of the parameter k.
• For W[2], a similar p-PCP characterization is derived with the same query and randomness complexity, showing that the framework generalizes beyond W[1].
• The class A[l] is shown to be in p-PCP with complexity depending on f(k), g(k), and the input size, with the bound reducing to the W[1] case when l=1.
• AW[*] is contained in a p-PCP with complexity scaling as f(k)n^{g(k)·⌊l/2⌋}·log(f(k)n^{O(1)})·log log(f(k)n^{O(1})), reflecting the number of even-indexed variable sets.
• The construction maintains correctness with error probability at most 1/2, and the verifier runs in time bounded by f(k)·|x|^{O(1)}.
• The results suggest that a tight p-PCP for W[1] could generalize to W[t] for all t, with t playing a central role in the complexity description.