English

Generalized Score Matching for General Domains

Methodology 2020-09-25 v1 Machine Learning

Abstract

Estimation of density functions supported on general domains arises when the data is naturally restricted to a proper subset of the real space. This problem is complicated by typically intractable normalizing constants. Score matching provides a powerful tool for estimating densities with such intractable normalizing constants, but as originally proposed is limited to densities on Rm\mathbb{R}^m and R+m\mathbb{R}_+^m. In this paper, we offer a natural generalization of score matching that accommodates densities supported on a very general class of domains. We apply the framework to truncated graphical and pairwise interaction models, and provide theoretical guarantees for the resulting estimators. We also generalize a recently proposed method from bounded to unbounded domains, and empirically demonstrate the advantages of our method.

Keywords

Cite

@article{arxiv.2009.11428,
  title  = {Generalized Score Matching for General Domains},
  author = {Shiqing Yu and Mathias Drton and Ali Shojaie},
  journal= {arXiv preprint arXiv:2009.11428},
  year   = {2020}
}

Comments

50 pages, 14 figures

R2 v1 2026-06-23T18:45:24.464Z