Efficient Symbolic Integration for Probabilistic Inference

Authors: Samuel Kolb, Martin Mladenov, Scott Sanner, Vaishak Belle, Kristian Kersting

IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical results demonstrate that these contributions can lead to a significant computational reduction over existing probabilistic inference algorithms.
Researcher Affiliation Collaboration a KU Leuven b Google Research c University of Toronto d University of Edinburgh e Alan Turing Institute f TU Darmstadt
Pseudocode Yes Algorithm 1 Integration using path enumeration. Algorithm 2 Integration using bound resolution (2a) Algorithm for bound-pair caching (2b) Algorithm for symbolic caching
Open Source Code Yes The code for the implementations used in our evaluation can be found at: https://xadd-wmi.github.io.
Open Datasets No The paper mentions generating synthetic problems and using a software package to generate problem instances but does not provide direct access or a formal citation for a specific publicly available dataset used in their experiments. For example: 'Using the publicly available code for the PA solver we generated 50 such WMI problems' and 'random pairs of nested SMT formulas and polynomial case functions in the form of Py SMT ASTs, generated using the predicate-abstraction solver software package.'
Dataset Splits No The paper discusses using 'synthetic benchmark problems' and generating '100 queries' for evaluation but does not provide specific percentages or counts for training, validation, or testing splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. It only mentions general terms without specific configurations.
Software Dependencies No The paper mentions 'Py SMT package [Gario and Micheli, 2015]' and 'sympy [Meurer et al., 2017]' but does not provide specific version numbers for these or other software dependencies, which are necessary for full reproducibility.
Experiment Setup Yes For every problem we record the execution time per solver, with a time-out of 60s.