[Paper Review] Exhaustive enumeration unveils clustering and freezing in random 3-SAT

John Ardelius, Lenka Zdeborová|arXiv (Cornell University)|Apr 2, 2008

Constraint Satisfaction and Optimization4 citations

TL;DR

This paper uses exhaustive enumeration of all solutions in random 3-SAT instances to study clustering and freezing phenomena, demonstrating strong agreement between analytical predictions and finite-size numerical results. It locates the freezing transition at α_f = 4.254 ± 0.009, closely matching the performance limit of state-of-the-art solvers, supporting the conjecture that freezing underlies algorithmic hardness.

ABSTRACT

We study geometrical properties of the complete set of solutions of the random 3-satisfiability problem. We show that even for moderate system sizes the number of clusters corresponds surprisingly well with the theoretic asymptotic prediction. We locate the freezing transition in the space of solutions which has been conjectured to be relevant in explaining the onset of computational hardness in random constraint satisfaction problems.

Motivation & Objective

• To investigate the relevance of asymptotic statistical physics predictions for finite-size random 3-SAT instances.
• To numerically verify the predicted number of solution clusters and compare it with analytical complexity functions.
• To locate the freezing transition in the solution space of random 3-SAT, a phenomenon conjectured to explain computational hardness.
• To assess whether the freezing transition correlates with the performance threshold of modern SAT solvers.
• To establish a robust, finite-size definition of clusters and frozen variables applicable to real-world and moderate-size instances.

Proposed method

• Exhaustive enumeration of all solutions for random 3-SAT instances up to N = 20 variables.
• Definition of clusters as connected components in a solution graph where solutions differ by a single variable flip.
• Application of the whitening procedure to identify frozen variables—those that remain fixed across all solutions in a cluster.
• Use of finite-size scaling to estimate the freezing transition point via crossing of unfrozen solution probabilities across system sizes.
• Cubic fitting of probability curves P_f(α, N) to locate the critical α where curves for different N intersect.
• Comparison of numerical results with analytical predictions from the cavity method and replica symmetry breaking.

Experimental results

Research questions

• RQ1 How well do asymptotic predictions for the number of solution clusters in random 3-SAT match numerical results on moderate-size instances?
• RQ2 At what constraint density α does the freezing transition occur in random 3-SAT, and how does it relate to the satisfiability threshold?
• RQ3 Does the location of the freezing transition coincide with the performance limit of the best-known SAT solvers?
• RQ4 How do finite-size effects influence the detection of clustering and freezing compared to the thermodynamic limit?
• RQ5 Can the whitening and cluster definition based on single-variable differences be used as a reliable proxy for theoretical cavity clusters in finite-size problems?

Key findings

• The number of solution clusters obtained via exhaustive enumeration shows excellent quantitative agreement with the analytical complexity function predicted by the cavity method for random 3-SAT.
• The freezing transition is located at α_f = 4.254 ± 0.009, which is extremely close to the satisfiability threshold α_s = 4.267.
• The freezing transition occurs at a constraint density that matches the performance limit of survey propagation and stochastic local search algorithms, supporting the hypothesis that freezing causes algorithmic hardness.
• The crossing point of P_f(α, N) curves for different system sizes shows minimal N-dependence, indicating robust convergence to the critical point.
• Clustering properties are less sensitive to finite-size effects than solution-level statistics, suggesting greater stability for cluster-based analysis.
• The proposed cluster and whitening definitions are applicable to any finite K-SAT instance and offer a practical framework for analyzing real-world and moderate-size problems.