A Compiler for Weak Decomposable Negation Normal Form

Authors: Petr Illner, Petr Kučera

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments demonstrate that w DNNF circuits are suitable for configuration instances. For the experiments, we considered the same data set... All the experiments were conducted on a Linux (Debian 11) machine using an AMD EPYC 7543 2.8GHz processor and 512 Gi B of RAM.
Researcher Affiliation Academia Petr Illner, Petr Kuˇcera Department of Theoretical Computer Science and Mathematical Logic, Faculty of Mathematics and Physics Charles University, Czech Republic illner3@gmail.com, kucerap@ktiml.mff.cuni.cz
Pseudocode Yes Algorithm 1: Bella(φ, cut)
Open Source Code Yes Bella (w DNNF circuits) and footnote 1https://github.com/Illner/Bella Compiler
Open Datasets Yes For the experiments, we considered the same data set7 which is used in (Lagniez and Marquis 2017). The data set, which contains 701 instances, consists of the following instance types: Bounded model checking (BMC), Bayesian networks (BN), circuit, configuration, handmade, planning, quantitative information flow analysis security (QIF), and random. and footnote 7http://www.cril.univ-artois.fr/kc/benchmarks.html
Dataset Splits No The paper mentions a dataset of 701 instances but does not provide specific train/validation/test dataset splits, percentages, or explicit sample counts for reproduction.
Hardware Specification Yes All the experiments were conducted on a Linux (Debian 11) machine using an AMD EPYC 7543 2.8GHz processor and 512 Gi B of RAM.
Software Dependencies Yes The same SAT solver as D42 (that is, a modified version of the SAT solver MINISAT 2.2 (E en and S orensson 2004) adapted for D4).
Experiment Setup Yes The time-out (resp. memory-out) was set to two hours (resp. 16 GB). Due to randomness, every instance was compiled five times, and the results presented in this section are averages.