Continual Learning in Generative Adversarial Nets

TLDR

Deep generative models now enable tractable learning of high‑dimensional data distributions, yet standard i.i.d. training assumptions and conditional variants that model multiple distributions in a single network are prone to catastrophic forgetting when distributions are encountered sequentially. The study aims to enable continual generative modeling by adapting recent catastrophic‑forgetting reduction techniques to the training of generative adversarial networks on a sequence of distinct distributions.

Abstract

Developments in deep generative models have allowed for tractable learning of high-dimensional data distributions. While the employed learning procedures typically assume that training data is drawn i.i.d. from the distribution of interest, it may be desirable to model distinct distributions which are observed sequentially, such as when different classes are encountered over time. Although conditional variations of deep generative models permit multiple distributions to be modeled by a single network in a disentangled fashion, they are susceptible to catastrophic forgetting when the distributions are encountered sequentially. In this paper, we adapt recent work in reducing catastrophic forgetting to the task of training generative adversarial networks on a sequence of distinct distributions, enabling continual generative modeling.