Latent Space Factorisation and Manipulation via Matrix Subspace Projection
This work addresses the need for simpler and more flexible latent space factorisation methods for researchers and practitioners in machine learning, though it is incremental as it builds on existing autoencoder techniques.
The paper tackles the problem of disentangling latent spaces in autoencoders to separate labeled attribute information from other characteristics, enabling attribute manipulation while preserving other information, with results showing competitive generation quality to strong baselines.
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 their 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.