Tractable Abstract Argumentation via Backdoor-Treewidth

Authors: Wolfgang Dvořák, Markus Hecher, Matthias König, André Schidler, Stefan Szeider, Stefan Woltran5608-5615

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

Reproducibility Variable Result LLM Response
Research Type Experimental We provide systems to find and exploit backdoors of small width, and conduct systematic experiments evaluating the new parameter.
Researcher Affiliation Academia TU Wien, Institute of Logic and Computation {dvorak, hecher, mkoenig, woltran}@dbai.tuwien.ac.at, {aschidler, sz}@ac.tuwien.ac.at
Pseudocode No The paper describes algorithmic steps and includes a flowchart (Figure 2), but it does not present any formal pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes We refer to the implemented system by arg BTW2. It can easily be extended to the credulous/skeptical acceptance-problem, the existence of extensions-problem, etc. ... 2Find arg BTW and benchmarks at github.com/mk-tu/arg BTW.
Open Datasets Yes As we focused on extreme values for backdoor size and treewidth, in addition to using instances provided by the ICCMA competition, we generated AFs to range from small to high backdoor sizes/treewidths. ... For scenario (B), we combined small ICCMA instances with the generated frameworks from (A), resulting in 1134 instances: (B1) 378 combinations of only ICCMA instances, (B2) 378 combinations of only generated instances, (B3) 378 combinations of ICCMA and generated instances.
Dataset Splits Yes For scenario (B), we combined small ICCMA instances with the generated frameworks from (A), resulting in 1134 instances: (B1) 378 combinations of only ICCMA instances, (B2) 378 combinations of only generated instances, (B3) 378 combinations of ICCMA and generated instances.
Hardware Specification Yes All our solvers ran on a cluster consisting of 12 servers. Each of these servers is equipped with two Intel Xeon E5-2650 CPUs, consisting of 12 physical cores that run at 2.2 GHz clock speed and have access to 256 GB shared main memory (RAM).
Software Dependencies Yes Results are gathered on Ubuntu 16.04.1 LTS powered on kernel 4.4.0-139 with hyperthreading disabled using version 3.7.6 of Python3. ... clingo3 (stable) and clingo (admissible): these configurations are based on clingo version 5.4.0 and we used arguments -q and -n 0 when counting answer sets.
Experiment Setup Yes In our experiments, we mainly compare wall clock time and number of solved instances over the best of two runs for every solver and instance, where each run is allowed to use up to 1200s of wall clock time and 16GB of main memory (RAM). ... For Step (1) we allotted at most 30 seconds and Step (2) is finished within a second per instance, due to the decomposer htd being reasonably fast.