Contextual Symmetries in Probabilistic Graphical Models

Authors: Ankit Anand, Aditya Grover, Mausam, Parag Singla

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on several domains of interest demonstrate that exploiting contextual symmetries can result in significant computational gains.Our experiments evaluate the use of contextual symmetries for faster inference in graphical models. We compare our approach against Orbital MCMC, which is the only available algorithm that exploits symmetries in a general MCMC framework. We also compare with vanilla Gibbs sampling, which does not exploit any symmetries.
Researcher Affiliation Academia Ankit Anand Indian Institute of Technology, Delhi ankit.anand@cse.iitd.ac.in Aditya Grover Stanford University adityag@cs.stanford.edu Mausam and Parag Singla Indian Institute of Technology, Delhi {mausam,parags}@cse.iitd.ac.in
Pseudocode No The paper describes the CON-MCMC algorithm in detail (Section 4) but does not present it in a structured pseudocode block or algorithm box.
Open Source Code Yes We also release a reference implementation of CON-MCMC sampler for wider use.1 1https://github.com/dair-iitd/con-mcmc
Open Datasets No The paper describes custom-built domains ('Sports Network' and 'Young and Old') but does not provide access information (link, DOI, or formal citation) for any publicly available or open datasets used in the experiments.
Dataset Splits No The paper discusses various experimental conditions and domain characteristics, but it does not provide specific dataset split information (e.g., percentages, sample counts, or citations to predefined splits) for training, validation, or testing.
Hardware Specification Yes All the experiments are run on a quad-core Intel i-7 processor.
Software Dependencies Yes We implement CONMCMC( ) as an extension of the original Orbital MCMC implementation3 available in the GAP language [GAP, 2015]. The existing implementation uses Saucy [Darga et al., 2008] for graph isomorphism and Gibbs sampler as the base Markov chain.
Experiment Setup Yes We show CON-MCMC results for = 0 and 0.01, which was chosen based on performance on smaller problem sizes.Figure 4 shows the performance of CON-MCMC( ) for different values of in the range 0.001 to 0.5 for both Sports network (single) and Y &O (single) domains.