Complex Markov Logic Networks: Expressivity and Liftability
Artificial Intelligence
2020-07-17 v2 Machine Learning
Logic in Computer Science
Abstract
We study expressivity of Markov logic networks (MLNs). We introduce complex MLNs, which use complex-valued weights, and we show that, unlike standard MLNs with real-valued weights, complex MLNs are fully expressive. We then observe that discrete Fourier transform can be computed using weighted first order model counting (WFOMC) with complex weights and use this observation to design an algorithm for computing relational marginal polytopes which needs substantially less calls to a WFOMC oracle than a recent algorithm.
Cite
@article{arxiv.2002.10259,
title = {Complex Markov Logic Networks: Expressivity and Liftability},
author = {Ondrej Kuzelka},
journal= {arXiv preprint arXiv:2002.10259},
year = {2020}
}
Comments
Fixed typos in Lemma 1 and Section 7. Paper accepted to UAI 2020