Pseudo-Rehearsal for Continual Learning with Normalizing Flows

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 conditioned on the task, 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.

A preliminary unpublished version of this work was presented in the LifelongML workshop, at ICML 2020

References in corpus (8)

• Bootstrap your own latent: A new approach to self-supervised Learning
• NICE: Non-linear Independent Components Estimation
• Normalizing Flows: An Introduction and Review of Current Methods
• Efficient Lifelong Learning with A-GEM
• Don't forget, there is more than forgetting: new metrics for Continual Learning
• Continual Unsupervised Representation Learning
• Learning Likelihoods with Conditional Normalizing Flows
• Efficient Continual Learning in Neural Networks with Embedding Regularization

Cited by in corpus (2)

• Task-agnostic Continual Learning with Hybrid Probabilistic Models
• Revisiting Catastrophic Forgetting in Class Incremental Learning