On Probabilistic Generalization of Backdoors in Boolean Satisfiability

Authors: Alexander Semenov, Artem Pavlenko, Daniil Chivilikhin, Stepan Kochemazov10353-10361

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Results of computational experiments show that the use of (ε, δ)-SBSes found by the proposed algorithm allows speeding up solving of test problems related to equivalence checking and hard crafted and combinatorial benchmarks compared to state-of-the-art SAT solvers.
Researcher Affiliation Academia Alexander Semenov, Artem Pavlenko, Daniil Chivilikhin, Stepan Kochemazov, ITMO University, St. Petersburg, Russia
Pseudocode No The paper describes algorithms in prose but does not include structured pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes The proposed approach was implemented in Python in the form of a multi-threaded application Evo Guess2, using Py SAT (Ignatiev, Morgado, and Marques-Silva 2018) for interfacing with SAT solvers. We used incremental SAT solvers that are available in Py SAT: namely, we mainly ran Glucose 3.0 (g3), Glucose 4.1 (g4), and Ca Di Ca L 1.0.3 (cd), though we also used Minisat 2.2 (m22) in one experiment. ... 2https://github.com/ctlab/EvoGuess/releases/tag/v2.0.0
Open Datasets Yes The first set of benchmarks belongs to the general class of equivalence checking instances (Kuehlmann and Krohm 1997; Molitor and Mohnke 2007). ... We also considered some crafted tests: sgen (Spence 2015, 2017), Pigeonhole Principle formulas (PHP), Parity principle (par9), and pmg12 from SAT competition.
Dataset Splits No The paper refers to using 'test problems' and 'benchmarks' but does not specify any training, validation, or test dataset splits or cross-validation methodology for their experiments.
Hardware Specification Yes Preliminary experiments were done using one node of the HPC-cluster Academician V.M. Matrosov 3 (with two 18-core Intel Xeon E5-2695 CPUs). For main experiments, we used one node of a computing cluster in ITMO University equipped with an Intel Xeon Gold 6248R CPU @ 3.00 GHz.
Software Dependencies Yes Py SAT (Ignatiev, Morgado, and Marques-Silva 2018) for interfacing with SAT solvers. We used incremental SAT solvers that are available in Py SAT: namely, we mainly ran Glucose 3.0 (g3), Glucose 4.1 (g4), and Ca Di Ca L 1.0.3 (cd), though we also used Minisat 2.2 (m22) in one experiment.
Experiment Setup Yes In the reported experiments we started the search from the empty backdoor (strategy 2), and initially used random samples of size of N = 4000... We also doubled the value of N if for the current value we had ρB = 1... In the experiments, we used Q = 8, H = 2. ... In our experiments we used ω|X| {15, 20}...