Online Estimation of SAT Solving Runtime

TL;DR

This paper introduces an online linear model approach to estimate SAT solving runtime, addressing challenges like backtracking and restarts, and improving solver selection and prediction accuracy.

Contribution

It proposes a novel online estimation method for SAT solving cost using a linear model trained during search, enhancing prediction and solver selection.

Findings

Effective runtime predictions on random and structured problems

Early restart predictions improve subsequent estimates

Cost estimations assist in solver portfolio selection

Abstract

We present an online method for estimating the cost of solving SAT problems. Modern SAT solvers present several challenges to estimate search cost including non-chronological backtracking, learning and restarts. Our method uses a linear model trained on data gathered at the start of search. We show the effectiveness of this method using random and structured problems. We demonstrate that predictions made in early restarts can be used to improve later predictions. We also show that we can use such cost estimations to select a solver from a portfolio.