A Caltech Library Service

Unsupervised Controllable Generation with Self-Training

Chrysos, Grigorios G. and Kossaifi, Jean and Yu, Zhiding and Anandkumar, Anima (2020) Unsupervised Controllable Generation with Self-Training. . (Unpublished)

[img] PDF - Submitted Version
See Usage Policy.


Use this Persistent URL to link to this item:


Recent generative adversarial networks (GANs) are able to generate impressive photo-realistic images. However, controllable generation with GANs remains a challenging research problem. Achieving controllable generation requires semantically interpretable and disentangled factors of variation. It is challenging to achieve this goal using simple fixed distributions such as Gaussian distribution. Instead, we propose an unsupervised framework to learn a distribution of latent codes that control the generator through self-training. Self-training provides an iterative feedback in the GAN training, from the discriminator to the generator, and progressively improves the proposal of the latent codes as training proceeds. The latent codes are sampled from a latent variable model that is learned in the feature space of the discriminator. We consider a normalized independent component analysis model and learn its parameters through tensor factorization of the higher-order moments. Our framework exhibits better disentanglement compared to other variants such as the variational autoencoder, and is able to discover semantically meaningful latent codes without any supervision. We demonstrate empirically on both cars and faces datasets that each group of elements in the learned code controls a mode of variation with a semantic meaning, e.g. pose or background change. We also demonstrate with quantitative metrics that our method generates better results compared to other approaches.

Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription Paper
Additional Information:Work done during GC’s intern at NVIDIA. We specially thank Weili Nie, Timo Aila and Sanja Fidler for the valuable discussions. We also thank the other AI-Algo team members for providing valuable feedback that improved this work.
Record Number:CaltechAUTHORS:20201106-120158552
Persistent URL:
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:106485
Deposited By: George Porter
Deposited On:06 Nov 2020 21:59
Last Modified:06 Nov 2020 21:59

Repository Staff Only: item control page