Learning Chordal Markov Networks via Branch and Bound

Authors: Kari Rantanen, Antti Hyttinen, Matti Järvisalo

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we show that the approach dominates in terms of running times a recent integer programming approach (and thereby also a recent constraint optimization approach) for the problem.
Researcher Affiliation Academia Kari Rantanen HIIT, Dept. Comp. Sci., University of Helsinki Antti Hyttinen HIIT, Dept. Comp. Sci., University of Helsinki Matti Järvisalo HIIT, Dept. Comp. Sci., University of Helsinki
Pseudocode Yes Algorithm 1 The core branch-and-bound search. Algorithm 2 Constructing parent set choices via dynamic programming. Algorithm 3 Computing upper bounds for a partial solution via dynamic programming.
Open Source Code No The paper does not provide concrete access to source code for the methodology described in this paper.
Open Datasets Yes We used a total of 54 real-world datasets used as standard benchmarks for exact approaches [32, 29].
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification Yes The experiments were run under Debian GNU/Linux on 2.83-GHz Intel Xeon E5440 nodes with 32-GB RAM.
Software Dependencies Yes We compare the performance of BBMarkov to that of GOBNILP (the newest development version [24] at the time of publication, using IBM CPLEX version 12.7.1 as the internal IP solver)
Experiment Setup Yes We did not impose a bound on the treewidth of the chordal graphs of interest, i.e., the size of candidate parent sets was not limited. We used the BDeu score with equivalent sample size 1. ... Figure 2 compares BBMarkov to GOBNILP and Junctor under a 1-h per-instance time limit... Our implementation uses d = 10.