Compilation of Aggregates in ASP Systems

Authors: Giuseppe Mazzotta, Francesco Ricca, Carmine Dodaro5834-5841

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

Reproducibility Variable Result LLM Response
Research Type Experimental We implement it on top of a state-of-the-art ASP system, and evaluate the performance on publicly-available benchmarks. Experiments show our approach is effective on ground-intensive ASP programs.
Researcher Affiliation Academia Giuseppe Mazzotta, Francesco Ricca, Carmine Dodaro Department of Mathematics and Computer Science, University of Calabria, Rende (CS), Italy giuseppe.mazzotta@unical.it, francesco.ricca@unical.it, carmine.dodaro@unical.it
Pseudocode Yes Algorithm 1 Compile Program; Algorithm 2 Compile Rule; Algorithm 3 Compile Rule With Sum; Algorithm 4 Compile Lazy Rule; Algorithm 5 Print Nested Join
Open Source Code No Finally, we mention that benchmarks and executables are publicly-available (Dodaro, Mazzotta, and Ricca 2021a,b).
Open Datasets Yes All benchmarks from ASP competitions (Calimeri et al. 2016) including at least one rule with aggregates that can be compiled under our conditions... Three grounding-intensive benchmarks taken from the literature, in particular Component Assignment Problem proposed by Alviano, Dodaro, and Maratea (2018), Dynamic In-Degree Counting and Exponential-Save proposed by Bomanson, Janhunen, and Weinzierl (2019).
Dataset Splits No The paper describes the use of various benchmarks and instances, but does not explicitly provide training, validation, or test dataset splits or specific methodologies for partitioning data.
Hardware Specification Yes Experiments were executed on Xeon(R) Gold 5118 CPUs running Ubuntu Linux (kernel 5.4.0-77-generic)
Software Dependencies Yes In all the experiments, the compilation-based approach, reported as WASP-COMP, has been compared with the plain version of WASP (Alviano et al. 2015) v. 169e40d and with the state-of-the-art system CLINGO v. 5.4.0 (Gebser et al. 2016).
Experiment Setup Yes time and memory are limited to 2100 seconds and 4GB, respectively.