Journal article
A Deep Generative Approach to Conditional Sampling
Journal of the American Statistical Association, Vol.118(543), pp.1837-1848
2023
DOI: 10.1080/01621459.2021.2016424
Abstract
We propose a deep generative approach to sampling from a conditional distribution based on a unified formulation of conditional distribution and generalized nonparametric regression function using the noise-outsourcing lemma. The proposed approach aims at learning a conditional generator, so that a random sample from the target conditional distribution can be obtained by transforming a sample drawn from a reference distribution. The conditional generator is estimated nonparametrically with neural networks by matching appropriate joint distributions using the Kullback-Liebler divergence. An appealing aspect of our method is that it allows either of or both the predictor and the response to be high-dimensional and can handle both continuous and discrete type predictors and responses. We show that the proposed method is consistent in the sense that the conditional generator converges in distribution to the underlying conditional distribution under mild conditions. Our numerical experiments with simulated and benchmark image data validate the proposed method and demonstrate that it outperforms several existing conditional density estimation methods.
Supplementary materials
for this article are available online.
Details
- Title: Subtitle
- A Deep Generative Approach to Conditional Sampling
- Creators
- Xingyu Zhou - University of IowaYuling Jiao - Wuhan UniversityJin Liu - Centre for Quantitative Medicine, Duke-NUS Medical School, Singapore;Jian Huang - University of Iowa
- Resource Type
- Journal article
- Publication Details
- Journal of the American Statistical Association, Vol.118(543), pp.1837-1848
- Publisher
- Taylor & Francis
- DOI
- 10.1080/01621459.2021.2016424
- ISSN
- 0162-1459
- eISSN
- 1537-274X
- Grant note
- DOI: 10.13039/100000001, name: the U.S. National Science Foundation, award: DMS-1916199; name: the National Science Foundation of China, award: 11871474; name: the Duke-NUS Medical School, award: R-913-200-098-263, MOE2018-T2-2-006; DOI: 10.13039/501100001459, name: Ministry of Education, Singapore; DOI: 10.13039/501100001809, name: National Natural Science Foundation of China; name: Duke-NUS Graduate Medical School WBS
- Language
- English
- Electronic publication date
- 02/03/2022
- Date published
- 2023
- Academic Unit
- Statistics and Actuarial Science
- Record Identifier
- 9984257621602771
Metrics
28 Record Views