Demystifying Inter-Class Disentanglement

Aviv Gabbay Yedid Hoshen

Abstract

Learning to disentangle the hidden factors of variations within a set of observations is a key task for artificial intelligence. We present a unified formulation for class and content disentanglement and use it to illustrate the limitations of current methods. We therefore introduce LORD, a novel method based on Latent Optimization for Representation Disentanglement. We find that latent optimization, along with an asymmetric noise regularization, is superior to amortized inference for achieving disentangled representations. In extensive experiments, our method is shown to achieve better disentanglement performance than both adversarial and non-adversarial methods that use the same level of supervision. We further introduce a clustering-based approach for extending our method for settings that exhibit in-class variation with promising results on the task of domain translation.

International Conference on Learning Representations (ICLR), 2020

Results (Content transfer between classes)

Cars3D

ML-VAE

DrNet

Ours

SmallNorb

ML-VAE

DrNet

Ours

KTH

ML-VAE

DrNet

Ours

CelebA

ML-VAE

DrNet

Ours

RaFD

Input

Angry

Contempt.

Disguste

Fearful

Happy

Sad

Surprised

Edges to Shoes

BibTeX

@inproceedings{gabbay2020lord,
  author    = {Aviv Gabbay and Yedid Hoshen},
  title     = {Demystifying Inter-Class Disentanglement},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2020}
}