Stochastic Constraint Propagation for Mining Probabilistic Networks
Authors: Anna Louise D. Latour, Behrouz Babaki, Siegfried Nijssen
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We experimentally evaluate the effectiveness of this global constraint in comparison to existing decomposition-based approaches, and show how this propagator can be used in combination with another data mining speciļ¬c constraint present in CP systems. As test cases we use problems from the data mining literature. 6 Experiments |
| Researcher Affiliation | Academia | Anna Louise D. Latour1 , Behrouz Babaki2 and Siegfried Nijssen3 1Leiden University, Leiden, The Netherlands 2Polytechnique Montr eal, Montreal, Canada 3UCLouvain, Louvain-la-Neuve, Belgium |
| Pseudocode | No | The paper describes algorithms (e.g., in Section 5.3 and 5.4) but does not present them in a structured pseudocode or algorithm block format. |
| Open Source Code | Yes | The implementations of our propagation algorithms and all the code for reproducing our experiments, are available at github.com/latower/SCMD. |
| Open Datasets | Yes | Test Data. Since we aim to improve on the performance of our earlier decomposition-based approach [Latour et al., 2017], we evaluate our methods on datasets used in this earlier work: DNA-protein interaction networks spine16, spine27a, spine27b [Ourfali et al., 2007] and collaboration networks datasets hep-th47 and hep-th5 [Kempe et al., 2003; Newman, 2001]. |
| Dataset Splits | No | The paper discusses using datasets (spine16, spine27a, spine27b, hep-th47, hep-th5) for evaluation but does not specify explicit training, validation, or test dataset splits (e.g., percentages, sample counts, or explicit splitting methodology). |
| Hardware Specification | Yes | Hardware. Our experiments ran on a machine with eight Xeon E5540 processors and 24GB RAM, under Cent OS Linux 7.4.1708. |
| Software Dependencies | Yes | We implemented the OBDD propagators proposed in Sections 5.3 and 5.4 in the Scala 2.12 library Osca R 4.0.0 [Osca R Team, 2012], because we need the state-of-the-art Cover Size 1.0.0 constraint [Schaus et al., 2017] for itemset mining to answer (Q5). We use CP solver Gecode 6.0.1 and MIP solver Gurobi 8.0.0 for comparison experiments. For modeling we use a SC-Prob Log version based on Prob Log 2.1 [DTAI Research Group, KU Leuven, 2015 2019], running in Python 3.6. We use the dd 0.5.4 library for the OBDD compilation. |
| Experiment Setup | Yes | Unless indicated otherwise, we use the default settings for all software. In our experiments to answer (Q1), we constrain both CP solvers to branch on the variables in lexicographical order. For the other experiments, our propagator uses the branching heuristic derivative1 (Section 5.4). |