Preferred Reasoning in ABA by Cycle-Breaking

Authors: Kiet Nguyen Anh, Markus Ulbricht

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

Reproducibility Variable Result LLM Response
Research Type Experimental We test our implementation against the ASPfor ABA solver which convincingly won the ABA track of the ICCMA 23 competition. As it turns out, our algorithm outperforms ASPfor ABA on instances with small backdoor sizes. and 5.2 Empirical Evaluation and Table 1: Runtimes of our Algorithm 1 and ASPfor ABA in s
Researcher Affiliation Academia Kiet Ngyuen Anh1 , Markus Ulbricht2 Sca DS.AI Dresden/Leipzig, Leipzig University 1kietnguyen2023@hotmail.com 2mulbricht@informatik.uni-leipzig.de and This word was funded by the Federal Ministry of Education and Research of Germany and by S achsische Staatsministerium f ur Wissenschaft, Kultur und Tourismus in the programme Center of Excellence for AI-research Center for Scalable Data Analytics and Artificial Intelligence Dresden/Leipzig , project identification number: Sca DS.AI.
Pseudocode Yes Algorithm 1 Computing Admissible Candidates
Open Source Code Yes The source code for our algorithm can be found online1. 1https://github.com/kiet Github User/ ABA-Backdoor-Implementation
Open Datasets Yes As experimental data, 400 ABAFs from the ICCMA 2023 competition benchmark generator2 were used. 2https://iccma2023.github.io/benchmarks.html
Dataset Splits No As experimental data, 400 ABAFs from the ICCMA 2023 competition benchmark generator2 were used.
Hardware Specification Yes Computations were done using resources of the Leipzig University Computing Center. There we used the Paul Cluster with a memory limit of 32GB ram and with the CPU: 2x AMD EPYC 7713 @ 2.0GHz Turbo 3.7GHz (64 cores).
Software Dependencies No We used the ASP-solver Clingo [Gebser et al., 2018] to compute a minimal ACYCDG-backdoor. Since we have to compute the deductive closure Th D(S) several times for different assumption sets S, we utilized the Glucose SAT-solver [Audemard and Simon, 2018] based on Mini Sat [E en and S orensson, 2004]. This SAT-solver is accessed through the Py SAT library [Ignatiev et al., 2018].
Experiment Setup No A timeout of 90s is used.