Continual Learning with Invertible Generative Models

TL;DR

This paper introduces a continual learning method using invertible generative models, specifically normalizing flows, to mitigate catastrophic forgetting with minimal additional memory and computational costs.

Contribution

It combines regularization and generative rehearsal in a novel way by using a single invertible neural network trained on internal embeddings.

Findings

Performs favorably against state-of-the-art methods

Maintains constant memory overhead due to invertibility

Offers bounded computational costs

Abstract

Catastrophic forgetting (CF) happens whenever a neural network overwrites past knowledge while being trained on new tasks. Common techniques to handle CF include regularization of the weights (using, e.g., their importance on past tasks), and rehearsal strategies, where the network is constantly re-trained on past data. Generative models have also been applied for the latter, in order to have endless sources of data. In this paper, we propose a novel method that combines the strengths of regularization and generative-based rehearsal approaches. Our generative model consists of a normalizing flow (NF), a probabilistic and invertible neural network, trained on the internal embeddings of the network. By keeping a single NF throughout the training process, we show that our memory overhead remains constant. In addition, exploiting the invertibility of the NF, we propose a simple approach to regularize the network's embeddings with respect to past tasks. We show that our method performs favorably with respect to state-of-the-art approaches in the literature, with bounded computational power and memory overheads.