Adversarial Learning of a Sampler Based on an Unnormalized Distribution

Chunyuan Li, Ke Bai, Jianqiao Li, Guoyin Wang, Changyou Chen, Lawrence Carin
Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, PMLR 89:3302-3311, 2019.

Abstract

Fundamental aspects of adversarial learning are investigated, with learning based on samples from the target distribution (conventional GAN setup). With insights so garnered, adversarial learning is extended to the case for which one has access to an unnormalized form $u(x)$ of the target density function, but no samples. Further, new concepts in GAN regularization are developed, based on learning from samples or from $u(x)$. The proposed method is compared to alternative approaches, with encouraging results demonstrated across a range of applications, including deep soft Q-learning.

Cite this Paper


BibTeX
@InProceedings{pmlr-v89-li19h, title = {Adversarial Learning of a Sampler Based on an Unnormalized Distribution}, author = {Li, Chunyuan and Bai, Ke and Li, Jianqiao and Wang, Guoyin and Chen, Changyou and Carin, Lawrence}, booktitle = {Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics}, pages = {3302--3311}, year = {2019}, editor = {Chaudhuri, Kamalika and Sugiyama, Masashi}, volume = {89}, series = {Proceedings of Machine Learning Research}, month = {16--18 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v89/li19h/li19h.pdf}, url = {http://proceedings.mlr.press/v89/li19h.html}, abstract = {Fundamental aspects of adversarial learning are investigated, with learning based on samples from the target distribution (conventional GAN setup). With insights so garnered, adversarial learning is extended to the case for which one has access to an unnormalized form $u(x)$ of the target density function, but no samples. Further, new concepts in GAN regularization are developed, based on learning from samples or from $u(x)$. The proposed method is compared to alternative approaches, with encouraging results demonstrated across a range of applications, including deep soft Q-learning.} }
Endnote
%0 Conference Paper %T Adversarial Learning of a Sampler Based on an Unnormalized Distribution %A Chunyuan Li %A Ke Bai %A Jianqiao Li %A Guoyin Wang %A Changyou Chen %A Lawrence Carin %B Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Masashi Sugiyama %F pmlr-v89-li19h %I PMLR %P 3302--3311 %U http://proceedings.mlr.press/v89/li19h.html %V 89 %X Fundamental aspects of adversarial learning are investigated, with learning based on samples from the target distribution (conventional GAN setup). With insights so garnered, adversarial learning is extended to the case for which one has access to an unnormalized form $u(x)$ of the target density function, but no samples. Further, new concepts in GAN regularization are developed, based on learning from samples or from $u(x)$. The proposed method is compared to alternative approaches, with encouraging results demonstrated across a range of applications, including deep soft Q-learning.
APA
Li, C., Bai, K., Li, J., Wang, G., Chen, C. & Carin, L.. (2019). Adversarial Learning of a Sampler Based on an Unnormalized Distribution. Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 89:3302-3311 Available from http://proceedings.mlr.press/v89/li19h.html.

Related Material