Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs

Authors: Mayukh Das, Devendra Singh Dhami, Gautam Kunapuli, Kristian Kersting, Sriraam Natarajan7816-7824

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results demonstrate the efficiency of these approximations, which can be applied to many complex statistical relational models, and can be significantly faster than state-of-the-art, both for inference and learning, without sacrificing effectiveness. We investigate the following questions: (Q1) Is MACH effective and efficient in full model learning with n-ary relations compared to a robust baseline? (Q2) Is modeling n-ary relations faithfully crucial when learning relational model? and (Q3) How does MACH compare (scaling vs. performance) to a state-of-the-art database-centric MLN system? ... Table 1 summarizes the performance and efficiency results of MACH against the baselines for structure and parameter learning of MLNs.
Researcher Affiliation Academia Mayukh Das University of Texas, Dallas mayukh.das1@utdallas.edu Devendra Singh Dhami University of Texas, Dallas devendra.dhami@utdallas.edu Gautam Kunapuli University of Texas, Dallas gautam.kunapuli@utdallas.edu Kristian Kersting Technical University of Darmstadt kersting@cs.tu-darmstadt.de Sriraam Natarajan University of Texas, Dallas sriraam.natarajan@utdallas.edu
Pseudocode Yes Algorithm 1 MACH: Motif-based Approximate Counting via Hypergraphs
Open Source Code Yes 3Code @ https://github.com/mayukhdas/MACH
Open Datasets Yes We used three standard SRL data sets: UW-CSE, Citeseer and Web KB, a biomedical data set Carcinogenesis (Srinivasan et al. 1997), and an NLP/Information Extraction(IE) data set NELL-Sports for evaluation.
Dataset Splits No We computed AUC-ROC, AUC-PR, CLL, F1 and running times averaged over 5 random train/test splits. The paper does not explicitly mention validation dataset splits.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running experiments.
Software Dependencies No The paper mentions 'Java-based Hypergraph DB architecture' but does not provide specific version numbers for Java or Hypergraph DB, or any other software dependencies.
Experiment Setup No The paper describes the system and baselines but does not provide specific experimental setup details such as hyperparameter values, model initialization, or training schedules.