AMP Chain Graphs: Minimal Separators and Structure Learning Algorithms

Authors: Mohammad Ali Javidian, Marco Valtorta, Pooyan Jamshidi

JAIR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using standard benchmarks and synthetically generated models and data in our experiments demonstrate the competitive performance of our decomposition-based method, called LCD-AMP, in comparison with the (modified versions of) PC-like algorithm.
Researcher Affiliation Academia Mohammad Ali Javidian EMAIL Marco Valtorta EMAIL Pooyan Jamshidi EMAIL Department of Computer Science and Engineering University of South Carolina, Columbia, SC, 29201, USA.
Pseudocode Yes Algorithm 1: Test for minimal separation (Problem 1) ... Algorithm 6: LCD-AMP : A decomposition-based recovery algorithm for AMP CGs
Open Source Code Yes Code for reproducing our results is available at https://github.com/majavid/AMPCGs2019.
Open Datasets Yes Using standard benchmarks and synthetically generated models and data in our experiments demonstrate the competitive performance of our decomposition-based method, called LCD-AMP... We perform simulation studies for four well-known Bayesian networks from Bayesian Network Repository (Scutari, 2017): ASIA, INSURANCE, ALARM, and HAILFINDER.
Dataset Splits No The paper describes generating i.i.d. samples of size n {500, 1000, 5000, 10000} for random AMP CGs and using 'observations' for discrete Bayesian networks. While sample sizes are given, there are no explicit details on training, validation, or testing splits.
Hardware Specification No The paper mentions parallel computations and distributing tasks over different cores, but it does not specify any particular CPU or GPU models, or other detailed hardware specifications for the experiments.
Software Dependencies No The paper mentions 'gaussCItest() function from the R package pcalg' and 'bnlearn R package (Scutari, 2017)'. While R packages are named, specific version numbers for these packages or for R itself are not provided.
Experiment Setup Yes In our simulation, we change three parameters p (the number of vertices), n (sample size) and N (expected number of adjacent vertices) as follows: p {10, 20, 30, 40, 50}, n {500, 1000, 5000, 10000}, and N {2, 3}. For each sample, three different significance levels (α = 0.005, 0.01, 0.05) are used to perform the hypothesis tests.