Latent Space Factorisation and Manipulation via Matrix Subspace Projection

Xiao Li*, Chenghua Lin, Ruizhe Li, Chaozheng Wang, Frank Guerin

*Corresponding author for this work

Research output: Contribution to journalConference article

1 Downloads (Pure)

Abstract

We tackle the problem disentangling the latent space of an autoencoder in order to separate labelled attribute information from other characteristic information. This then allows us to change selected attributes while preserving other information. Our method, matrix subspace projection, is much simpler than previous approaches to latent space factorisation, for example not requiring multiple discriminators or a careful weighting among loss functions. Furthermore our new model can be applied to autoencoders as a plugin, and works across diverse domains such as images or text. We demonstrate the utility of our method for attribute manipulation in autoencoders trained across varied domains, using both human evaluation and automated methods. The quality of generation of our new model (e.g. reconstruction, conditional generation) is highly competitive to a number of strong baselines.

Original languageEnglish
Number of pages11
JournalProceedings of Machine Learning Research
Volume119
Publication statusAccepted/In press - 16 Jun 2020
Event37th International Conference on Machine Learning - Vienna, Austria
Duration: 12 Jul 202018 Jul 2020
https://icml.cc/Conferences/2020/CallForPapers

Fingerprint Dive into the research topics of 'Latent Space Factorisation and Manipulation via Matrix Subspace Projection'. Together they form a unique fingerprint.

  • Cite this