MaskAAE: Latent space optimization for Adversarial Auto-Encoders

Arnab Mondal, Sankalan Pal Chowdhury, Aravind Jayendran, Himanshu Asnani, Parag Singla, Prathosh A P
Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI), PMLR 124:689-698, 2020.

Abstract

The field of neural generative models is dominated by the highly successful Generative Adversarial Networks (GANs) despite their challenges, such as training instability and mode collapse. Auto-Encoders (AE) with regularized latent space provide an alternative framework for generative models, albeit their performance levels have not reached that of GANs. In this work, we hypothesise that the dimensionality of the AE model’s latent space has a critical effect on the quality of generated data. Under the assumption that nature generates data by sampling from a “true" generative latent space followed by a deterministic function, we show that the optimal performance is obtained when the dimensionality of the latent space of the AE-model matches with that of the “true" generative latent space. Further, we propose an algorithm called the Mask Adversarial Auto-Encoder (MaskAAE), in which the dimensionality of the latent space of an adversarial auto encoder is brought closer to that of the “true" generative latent space, via a procedure to mask the spurious latent dimensions. We demonstrate through experiments on synthetic and several real-world datasets that the proposed formulation yields betterment in the generation quality.

Cite this Paper


BibTeX
@InProceedings{pmlr-v124-mondal20a, title = {MaskAAE: Latent space optimization for Adversarial Auto-Encoders}, author = {Mondal, Arnab and Pal Chowdhury, Sankalan and Jayendran, Aravind and Asnani, Himanshu and Singla, Parag and A P, Prathosh}, booktitle = {Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI)}, pages = {689--698}, year = {2020}, editor = {Jonas Peters and David Sontag}, volume = {124}, series = {Proceedings of Machine Learning Research}, month = {03--06 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v124/mondal20a/mondal20a.pdf}, url = { http://proceedings.mlr.press/v124/mondal20a.html }, abstract = {The field of neural generative models is dominated by the highly successful Generative Adversarial Networks (GANs) despite their challenges, such as training instability and mode collapse. Auto-Encoders (AE) with regularized latent space provide an alternative framework for generative models, albeit their performance levels have not reached that of GANs. In this work, we hypothesise that the dimensionality of the AE model’s latent space has a critical effect on the quality of generated data. Under the assumption that nature generates data by sampling from a “true" generative latent space followed by a deterministic function, we show that the optimal performance is obtained when the dimensionality of the latent space of the AE-model matches with that of the “true" generative latent space. Further, we propose an algorithm called the Mask Adversarial Auto-Encoder (MaskAAE), in which the dimensionality of the latent space of an adversarial auto encoder is brought closer to that of the “true" generative latent space, via a procedure to mask the spurious latent dimensions. We demonstrate through experiments on synthetic and several real-world datasets that the proposed formulation yields betterment in the generation quality.} }
Endnote
%0 Conference Paper %T MaskAAE: Latent space optimization for Adversarial Auto-Encoders %A Arnab Mondal %A Sankalan Pal Chowdhury %A Aravind Jayendran %A Himanshu Asnani %A Parag Singla %A Prathosh A P %B Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI) %C Proceedings of Machine Learning Research %D 2020 %E Jonas Peters %E David Sontag %F pmlr-v124-mondal20a %I PMLR %P 689--698 %U http://proceedings.mlr.press/v124/mondal20a.html %V 124 %X The field of neural generative models is dominated by the highly successful Generative Adversarial Networks (GANs) despite their challenges, such as training instability and mode collapse. Auto-Encoders (AE) with regularized latent space provide an alternative framework for generative models, albeit their performance levels have not reached that of GANs. In this work, we hypothesise that the dimensionality of the AE model’s latent space has a critical effect on the quality of generated data. Under the assumption that nature generates data by sampling from a “true" generative latent space followed by a deterministic function, we show that the optimal performance is obtained when the dimensionality of the latent space of the AE-model matches with that of the “true" generative latent space. Further, we propose an algorithm called the Mask Adversarial Auto-Encoder (MaskAAE), in which the dimensionality of the latent space of an adversarial auto encoder is brought closer to that of the “true" generative latent space, via a procedure to mask the spurious latent dimensions. We demonstrate through experiments on synthetic and several real-world datasets that the proposed formulation yields betterment in the generation quality.
APA
Mondal, A., Pal Chowdhury, S., Jayendran, A., Asnani, H., Singla, P. & A P, P.. (2020). MaskAAE: Latent space optimization for Adversarial Auto-Encoders. Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI), in Proceedings of Machine Learning Research 124:689-698 Available from http://proceedings.mlr.press/v124/mondal20a.html .

Related Material