Fisher Auto-Encoders

TL;DR

Fisher auto-encoders are introduced as a robust alternative to traditional VAEs, minimizing Fisher divergence to better quantify model uncertainty and improve robustness in generative modeling.

Contribution

This paper proposes Fisher auto-encoders that utilize Fisher divergence for training, offering a new approach that more accurately measures the distance between true and modeled posteriors.

Findings

Fisher AEs outperform VAEs and Wasserstein AEs in robustness.

Fisher AEs provide exact quantification of posterior distribution differences.

Results on MNIST and celebA datasets demonstrate competitive performance.

Abstract

It has been conjectured that the Fisher divergence is more robust to model uncertainty than the conventional Kullback-Leibler (KL) divergence. This motivates the design of a new class of robust generative auto-encoders (AE) referred to as Fisher auto-encoders. Our approach is to design Fisher AEs by minimizing the Fisher divergence between the intractable joint distribution of observed data and latent variables, with that of the postulated/modeled joint distribution. In contrast to KL-based variational AEs (VAEs), the Fisher AE can exactly quantify the distance between the true and the model-based posterior distributions. Qualitative and quantitative results are provided on both MNIST and celebA datasets demonstrating the competitive performance of Fisher AEs in terms of robustness compared to other AEs such as VAEs and Wasserstein AEs.