Imputation estimators for unnormalized models with missing data

Thumbnail Image
Date
2019-03-12
Authors
Uehara, Masatoshi
Matsuda, Takeru
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract

We propose estimation methods for unnormalized models with missing data. The key concept is to combine a modern imputation technique with estimators for unnormalized models including noise contrastive estimation and score matching. Further, we derive asymptotic distributions of the proposed estimators and construct the confidence intervals. The application to truncated Gaussian graphical models with missing data shows the validity of the proposed methods.

Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Type
article
Comments

This pre-print is made available through arxiv: https://arxiv.org/abs/1903.03630.

Rights Statement
Copyright
Tue Jan 01 00:00:00 UTC 2019
Funding
DOI
Supplemental Resources
Source
Collections