Single-Shot Epistemic Logic Program Solving
Authors: Manuel Bichler, Michael Morak, Stefan Woltran
IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experimental Evaluation We tested our system selp against the state-of-the-art ELP solver, EP-ASP [Son et al., 2017], using three test sets. For every test set, we measured the time it took to solve the consistency problem. |
| Researcher Affiliation | Academia | Manuel Bichler, Michael Morak, and Stefan Woltran TU Wien, Vienna, Austria {bichler,morak,woltran}@dbai.tuwien.ac.at |
| Pseudocode | No | The paper describes program constructions and rule forms but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | We implemented the reduction in the previous section as part of the single shot ELP solving toolbox selp, available at dbai.tuwien.ac.at/proj/selp. |
| Open Datasets | Yes | Benchmark Instances. We used three types of benchmarks, two coming from the ELP literature and one from the QSAT domain that contains structures of low treewidth5. Scholarship Eligibility (SE). This set of non-ground ELP programs is shipped together with EP-ASP. ... Tree QBFs (TQ). The hardness proof for ELP consistency [Shen and Eiter, 2016] relies on a reduction from certain QBFs with three quantifier blocks. We apply this reduction to the 14 Tree instances of QBFEVAL 16 [Pulina, 2016], available at http://www.qbflib.org/ family_detail.php?id Family=56, splitting each instance s variables into three random quantifier blocks. Benchmark archive: dbai.tuwien.ac.at/proj/selp |
| Dataset Splits | No | The paper mentions using 'test sets' and 'benchmark instances' for evaluation but does not specify training, validation, or test dataset splits needed for reproduction of the model's training process. |
| Hardware Specification | Yes | Experiments were run on a 2.1GHz AMD Opteron 6272 system with 224 GB of memory. Each process was assigned a maximum of 14 GB of RAM. |
| Software Dependencies | Yes | For EP-ASP, we used the required clingo 4.5.3, since newer versions are incompatible with the solver. For selp, we used clingo 5.2.2, htd_main 1.2.0, and lpopt 2.2. |
| Experiment Setup | Yes | For selp, clasp was stopped after finding the first answer set. For EP-ASP, search was terminated after finding the first candidate world view4. ... EP-ASP was called with the preprocessing option for brave and cautious consequences on, since it always ran faster this way. ... The --sat-prepro=3 flag is recommended by lpopt |