NashAE: Disentangling Representations through Adversarial Covariance Minimization

TL;DR

NashAE is a self-supervised autoencoder-based method that disentangles data factors without prior knowledge by using a minmax game between the encoder and regression networks, improving reliability and capacity.

Contribution

It introduces NashAE, a novel adversarial covariance minimization approach for disentangling representations without prior assumptions.

Findings

Outperforms existing methods on disentanglement metrics

Increases reliability of learned representations

Captures salient data characteristics effectively

Abstract

We present a self-supervised method to disentangle factors of variation in high-dimensional data that does not rely on prior knowledge of the underlying variation profile (e.g., no assumptions on the number or distribution of the individual latent variables to be extracted). In this method which we call NashAE, high-dimensional feature disentanglement is accomplished in the low-dimensional latent space of a standard autoencoder (AE) by promoting the discrepancy between each encoding element and information of the element recovered from all other encoding elements. Disentanglement is promoted efficiently by framing this as a minmax game between the AE and an ensemble of regression networks which each provide an estimate of an element conditioned on an observation of all other elements. We quantitatively compare our approach with leading disentanglement methods using existing disentanglement metrics. Furthermore, we show that NashAE has increased reliability and increased capacity to capture salient data characteristics in the learned latent representation.