Finding All Bayesian Network Structures within a Factor of Optimal
Authors: Zhenyu A. Liao, Charupriya Sharma, James Cussens, Peter van Beek7892-7899
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental Evaluation In this section, we evaluate the proposed BF based method and compare its performance with published k-best solvers. |
| Researcher Affiliation | Academia | 1David R. Cheriton School of Computer Science, University of Waterloo, Waterloo, ON, Canada 2Department of Computer Science, University of York, York, United Kingdom |
| Pseudocode | No | The paper describes methods and rules but does not present any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The modified code is available at: https://www.cs.york.ac.uk/aig/sw/gobnilp/ |
| Open Datasets | Yes | The datasets are obtained from the UCI Machine Learning Repository (Dheeru and Karra Taniskidou 2017) and the Bayesian Network Repository. |
| Dataset Splits | No | The paper does not explicitly provide details about training, validation, or test dataset splits. It mentions 'N' as the number of instances in the dataset, but no specific split percentages or counts. |
| Hardware Specification | Yes | All experiments are conducted on computers with 2.2 GHz Intel E7-4850V3 processors. Each experiment is limited to 64 GB of memory and 24 hours of CPU time. |
| Software Dependencies | Yes | The code is compiled with SCIP 6.0.0 and CPLEX 12.8.0. |
| Experiment Setup | Yes | We modified the development version (9c9f3e6) of GOBNILP, referred below as GOBNILP dev, to apply pruning rules presented above during scoring and supplied appropriate parameter settings for collecting near-optimal networks4. The code is compiled with SCIP 6.0.0 and CPLEX 12.8.0. ... To find only solutions with objective no worse than (OPT + ϵ), SCIP’s SCIPset Objlimit function is used. ... Finally the collected networks are categorized into Markov equivalence classes (MECs), where two networks belong to the same MEC iff they have the same skeleton and v-structures (Verma and Pearl 1990). The proposed approach is tested on datasets with up to 57 variables. The search time T, the number of collected networks |G| and the number of MECs M in the collected networks at BF = 3, 20 and 150 using BIC are reported in Table 1, where n is the number of random variables in the dataset and N is the number of instances in the dataset. |