Optimal Status Enforcement in Abstract Argumentation
Authors: Andreas Niskanen, Johannes P. Wallner, Matti Järvisalo
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We have implemented the Max SAT encodings and the CEGAR-procedures, obtaining the first system for optimal status enforcement. Here we present an overview of an empirical evaluation of the system. We generated benchmark instances following essentially a standard model for random directed graphs. Mean runtimes with timeouts included as 900s are shown in Figure 2 for the NP problems of credulous status enforcement with |N| under admissible semantics (left) and for the P2 skeptical and credulous status enforcement problems under stable semantics (right). |
| Researcher Affiliation | Academia | Andreas Niskanen and Johannes P. Wallner and Matti Järvisalo Helsinki Institute for Information Technology HIIT, Department of Computer Science, University of Helsinki, Finland |
| Pseudocode | Yes | Algorithm 1 CEGAR-based status enforcement for AF F = (A, R), P, N A, σ 2 {adm, stb}, M 2 {cred, skept} |
| Open Source Code | Yes | Our status enforcement system implementation together with benchmarks used in this paper, as well as full formal proofs of our complexity results, are available via http://www.cs.helsinki.fi/group/coreo/pakota/. |
| Open Datasets | No | No specific publicly available dataset with concrete access information (link, DOI, formal citation with authors/year) is provided. The paper states: 'We generated benchmark instances following essentially a standard model for random directed graphs. For each |A| 2 {20, 40, . . . , 200} and p 2 {0.05, 0.1, . . . , 0.35}2, we generated ten random AFs with |A| arguments by including individual attacks with probability p.' While their generated benchmarks are available via a general project URL mentioned earlier, this question refers to a dataset being publicly available with a formal citation/access, not the generated instances of their experiment. |
| Dataset Splits | No | No explicit details about training, validation, or test splits were provided. The paper describes how benchmark instances were generated: 'For each |A| 2 {20, 40, . . . , 200} and p 2 {0.05, 0.1, . . . , 0.35}2, we generated ten random AFs with |A| arguments by including individual attacks with probability p. For each AF, we randomly picked 5 arguments, of which we enforced |P| 2 {1, 2, . . . , 5} positively, and finally picked |N| 2 {0, 1, 2, 5} arguments from the set A \ P to be enforced negatively.' |
| Hardware Specification | Yes | We used Open WBO [Martins et al., 2014] as the Max SAT solver, and ran the experiments on 2.83-GHz Intel Xeon E5440 4-core nodes with 32-GB RAM and Debian GNU/Linux 8 under 900-second per-instance timeout. |
| Software Dependencies | No | The paper states: 'We used Open WBO [Martins et al., 2014] as the Max SAT solver, and ran the experiments on... Debian GNU/Linux 8'. While Open WBO is named, its version number is not provided, and Debian GNU/Linux is an operating system, not an ancillary software dependency with a specific version number to ensure reproducibility of the experimental software stack. |
| Experiment Setup | Yes | For each |A| 2 {20, 40, . . . , 200} and p 2 {0.05, 0.1, . . . , 0.35}2, we generated ten random AFs with |A| arguments by including individual attacks with probability p. For each AF, we randomly picked 5 arguments, of which we enforced |P| 2 {1, 2, . . . , 5} positively, and finally picked |N| 2 {0, 1, 2, 5} arguments from the set A \ P to be enforced negatively. We used Open WBO [Martins et al., 2014] as the Max SAT solver, and ran the experiments on 2.83-GHz Intel Xeon E5440 4-core nodes with 32-GB RAM and Debian GNU/Linux 8 under 900-second per-instance timeout. |