Guiding CDCL SAT Search via Random Exploration amid Conflict Depression

Authors: Md Solimul Chowdhury, Martin Müller, Jia You1428-1435

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental An extensive empirical evaluation with four state-of-the-art CDCL SAT solvers demonstrates good-to-strong performance gains with the exp SAT approach.
Researcher Affiliation Academia Md Solimul Chowdhury, Martin M uller, Jia-Huai You Department of Computing Science, University of Alberta Edmonton, Alberta, Canada. {mdsolimu, mmueller, jyou}@ualberta.ca
Pseudocode Yes Algorithm 1: Exploration Based CDCL Solver : exp SAT; Algorithm 2: Decide the Branching Variable
Open Source Code Yes Source code of the exp SAT solvers and 52 SATCoin instances are available at: https://figshare.com/articles/exp SAT Solvers Instances/10324172.
Open Datasets Yes Test Set 1 contains 750 instances from the main track of SAT-2017 (350) and 2018 (400) and is run with a time limit of 5000 seconds per instance. Test Set 2 consists of 52 hard instances from SATCoin (Bit Coin Mining) cryptographic benchmark, which are generated with the instance generator from (Manthey and Heusser 2018).
Dataset Splits No The paper describes test sets used for evaluation of SAT solvers but does not specify training, validation, or test splits in the context of training a machine learning model, as the research focuses on SAT solver extensions.
Hardware Specification Yes All experiments presented in this paper were run on a workstation with 64GB RAM and a processor clock speed of 2.4 GHz.
Software Dependencies Yes We adopted four baseline solvers: g LCM4, Mpl COMSPS5 (winner of SAT-2016), Mpl CM4 (second runner up of SAT-2018) and Mpl CBT4 (winner of SAT-2018). glucose 4.2.1, which implements the Learned Clause Minimization (LCM) (Luo et al. 2017) technique on top of glucose 4.1.
Experiment Setup Yes To set the values of the exploration parameters, we performed a small scale grid search with e GLCM: we took one instance at random out of each benchmark from SAT-2018, which gave a subset of 23 instances. We run e GLCM on this subset for small parameters ranges, l W and n W in [4,5,6] and pexp in [0.01,0.02,0.03]. From this grid search, we chose our default parameter setting (m W, m S, pexp)=(5, 5, 0.02). We set the value of the exponential decay parameter ω to 0.9 based on intuition. These are the values used in the experiments. The values of the other parameters for adaptive version are given in Table 7: Weights (w1, w2, w3) (40, 10, 3), Range for n W [ln W , un W ] [1, 20], Range for l W [ll W , ul W ] [1, 10], Range for pexp [lpexp, upexp] [0.02, 0.6], Step size for parameters (sn W , sl W , spexp) (1, 1, 0.01), Exponential Decay Factor ω 0.9.