Presentation | 2020-01-24 Optimal Transport based Autoencoder for class and style Disentanglement Florian Tambon, Tetsuo Furukawa, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | The Sinkhorn autoencoder is a novel generative model using optimal transport to model the aggregated posterior from samples, hence discarding traditional reparametrization trick from classical Variational Autoencoder (VAE) and allowing better flexibility of metrics spaces and priors. Yet, one of the down side of all latent space modelling methods is the lack of interpretability and the potential entanglement problem. The aim of this work is to extend the Sinkhorn Autoencoder to better disentangle the latent space by focusing on the class/style separation approach while providing better interpretability and generative capability. Thus, our method would help further expand knowledge regarding optimal transport based generative model. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Generative model / Optimal Transport / Disentanglement / Sinkhorn Loss / Autoencoder / Latent Space |
Paper # | NC2019-62 |
Date of Issue | 2020-01-16 (NC) |
Conference Information | |
Committee | NLP / NC |
---|---|
Conference Date | 2020/1/23(3days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Miyakojima Marine Terminal |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Hiroaki Kurokawa(Tokyo Univ. of Tech.) / Hayaru Shouno(UEC) |
Vice Chair | Kiyohisa Natsume(Kyushu Inst. of Tech.) / Kazuyuki Samejima(Tamagawa Univ) |
Secretary | Kiyohisa Natsume(Nippon Inst. of Tech.) / Kazuyuki Samejima(Kyushu Inst. of Tech.) |
Assistant | Yutaka Shimada(Saitama Univ.) / Toshikaza Samura(Yamaguchi Univ.) / Takashi Shinozaki(NICT) / Ken Takiyama(TUAT) |
Paper Information | |
Registration To | Technical Committee on Nonlinear Problems / Technical Committee on Neurocomputing |
---|---|
Language | ENG |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Optimal Transport based Autoencoder for class and style Disentanglement |
Sub Title (in English) | |
Keyword(1) | Generative model |
Keyword(2) | Optimal Transport |
Keyword(3) | Disentanglement |
Keyword(4) | Sinkhorn Loss |
Keyword(5) | Autoencoder |
Keyword(6) | Latent Space |
1st Author's Name | Florian Tambon |
1st Author's Affiliation | Kyushu Institute of Technology(Kyutech) |
2nd Author's Name | Tetsuo Furukawa |
2nd Author's Affiliation | Kyushu Institute of Technology(Kyutech) |
Date | 2020-01-24 |
Paper # | NC2019-62 |
Volume (vol) | vol.119 |
Number (no) | NC-382 |
Page | pp.pp.17-22(NC), |
#Pages | 6 |
Date of Issue | 2020-01-16 (NC) |