Perturbative methods for mostly monotonic probabilistic satisfiability problems

TL;DR

This paper extends perturbative methods from statistical physics to compute bounds on probabilistic satisfiability problems, especially focusing on monotonic cases, improving approximation accuracy.

Contribution

It introduces a novel approach combining weak- and strong-coupling methods to produce tight bounds on monotonic probabilistic satisfiability problems.

Findings

Bounds are tight and saturated by specific problem instances.

Extension of physics-based methods to heterogeneous satisfiability problems.

Provides a systematic way to estimate approximation errors.

Abstract

The probabilistic satisfiability of a logical expression is a fundamental concept known as the partition function in statistical physics and field theory, an evaluation of a related graph's Tutte polynomial in mathematics, and the Moore-Shannon network reliability of that graph in engineering. It is the crucial element for decision-making under uncertainty. Not surprisingly, it is provably hard to compute exactly or even to approximate. Many of these applications are concerned only with a subset of problems for which the solutions are monotonic functions. Here we extend the weak- and strong-coupling methods of statistical physics to heterogeneous satisfiability problems and introduce a novel approach to constructing lower and upper bounds on the approximation error for monotonic problems. These bounds combine information from both perturbative analyses to produce bounds that are tight in the sense that they are saturated by some problem instance that is compatible with all the information contained in either approximation.