AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE

Lu Changjie, Zheng Shen, Wang Zirui, Dib Omar, Gupta Gaurav
Proceedings of The 14th Asian Conference on Machine Learning, PMLR 189:658-673, 2023.

Abstract

Recently, introspective models like IntroVAE and S-IntroVAE have excelled in image generation and reconstruction tasks. The principal characteristic of introspective models is the adversarial learning of VAE, where the encoder attempts to distinguish between the real and the fake (i.e., synthesized) images. However, due to the unavailability of an effective metric to evaluate the difference between the real and the fake images, the posterior collapse and the vanishing gradient problem still exist, reducing the fidelity of the synthesized images. In this paper, we propose a new variation of IntroVAE called Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE). We theoretically analyze the vanishing gradient problem and construct a new Adversarial Similarity Distance (AS-Distance) using the 2-Wasserstein distance and the kernel trick. With weight annealing on AS-Distance and KL-Divergence, the AS-IntroVAE are able to generate stable and high-quality images. The posterior collapse problem is addressed by making per-batch attempts to transform the image so that it better fits the prior distribution in the latent space. Compared with the per-image approach, this strategy fosters more diverse distributions in the latent space, allowing our model to produce images of great diversity. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of AS-IntroVAE on image generation and reconstruction tasks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v189-changjie23a, title = {AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE}, author = {Changjie, Lu and Shen, Zheng and Zirui, Wang and Omar, Dib and Gaurav, Gupta}, booktitle = {Proceedings of The 14th Asian Conference on Machine Learning}, pages = {658--673}, year = {2023}, editor = {Khan, Emtiyaz and Gonen, Mehmet}, volume = {189}, series = {Proceedings of Machine Learning Research}, month = {12--14 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v189/changjie23a/changjie23a.pdf}, url = {https://proceedings.mlr.press/v189/changjie23a.html}, abstract = {Recently, introspective models like IntroVAE and S-IntroVAE have excelled in image generation and reconstruction tasks. The principal characteristic of introspective models is the adversarial learning of VAE, where the encoder attempts to distinguish between the real and the fake (i.e., synthesized) images. However, due to the unavailability of an effective metric to evaluate the difference between the real and the fake images, the posterior collapse and the vanishing gradient problem still exist, reducing the fidelity of the synthesized images. In this paper, we propose a new variation of IntroVAE called Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE). We theoretically analyze the vanishing gradient problem and construct a new Adversarial Similarity Distance (AS-Distance) using the 2-Wasserstein distance and the kernel trick. With weight annealing on AS-Distance and KL-Divergence, the AS-IntroVAE are able to generate stable and high-quality images. The posterior collapse problem is addressed by making per-batch attempts to transform the image so that it better fits the prior distribution in the latent space. Compared with the per-image approach, this strategy fosters more diverse distributions in the latent space, allowing our model to produce images of great diversity. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of AS-IntroVAE on image generation and reconstruction tasks.} }
Endnote
%0 Conference Paper %T AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE %A Lu Changjie %A Zheng Shen %A Wang Zirui %A Dib Omar %A Gupta Gaurav %B Proceedings of The 14th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Emtiyaz Khan %E Mehmet Gonen %F pmlr-v189-changjie23a %I PMLR %P 658--673 %U https://proceedings.mlr.press/v189/changjie23a.html %V 189 %X Recently, introspective models like IntroVAE and S-IntroVAE have excelled in image generation and reconstruction tasks. The principal characteristic of introspective models is the adversarial learning of VAE, where the encoder attempts to distinguish between the real and the fake (i.e., synthesized) images. However, due to the unavailability of an effective metric to evaluate the difference between the real and the fake images, the posterior collapse and the vanishing gradient problem still exist, reducing the fidelity of the synthesized images. In this paper, we propose a new variation of IntroVAE called Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE). We theoretically analyze the vanishing gradient problem and construct a new Adversarial Similarity Distance (AS-Distance) using the 2-Wasserstein distance and the kernel trick. With weight annealing on AS-Distance and KL-Divergence, the AS-IntroVAE are able to generate stable and high-quality images. The posterior collapse problem is addressed by making per-batch attempts to transform the image so that it better fits the prior distribution in the latent space. Compared with the per-image approach, this strategy fosters more diverse distributions in the latent space, allowing our model to produce images of great diversity. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of AS-IntroVAE on image generation and reconstruction tasks.
APA
Changjie, L., Shen, Z., Zirui, W., Omar, D. & Gaurav, G.. (2023). AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE. Proceedings of The 14th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 189:658-673 Available from https://proceedings.mlr.press/v189/changjie23a.html.

Related Material