Pushing Forward Marginal MAP with Best-First Search

Authors: Radu Marinescu, Rina Dechter, Alexander Ihler

IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We analyze the potential relative benefits of several bestfirst search algorithms and demonstrate their effectiveness against recent branch and bound schemes through extensive empirical evaluations. Our results show that best-first search improves significantly over existing depth-first approaches, in many cases by several orders of magnitude, especially when guided by relatively weak heuristics.
Researcher Affiliation Collaboration Radu Marinescu IBM Research Ireland radu.marinescu@ie.ibm.com Rina Dechter and Alexander Ihler University of California, Irvine Irvine, CA 92697, USA {dechter,ihler}@ics.uci.edu
Pseudocode Yes Algorithm 1: AOBF-MMAP
Open Source Code No No statement or link is provided indicating the release of the source code for the methodology.
Open Datasets Yes We evaluate empirically the proposed best-first search algorithms on problem instances derived from benchmarks used in the PASCAL2 Inference Challenge [Elidan et al., 2012].
Dataset Splits No The paper describes generating 'easy' and 'hard' problem instances for evaluation, but does not specify traditional training, validation, or test splits of a dataset as would be typical for machine learning model training.
Hardware Specification Yes All algorithms were implemented in C++ (64-bit) and the experiments were run on a 2.6GHz 8-core processor with 80 GB of RAM.
Software Dependencies No All algorithms were implemented in C++ (64-bit)
Experiment Setup Yes The weighted mini-bucket heuristics were generated in a preprocessing phase, prior to search, using uniform weights.