Class-Incremental Learning with Generative Classifiers

TL;DR

This paper introduces a novel class-incremental learning approach called generative classification, which models the joint distribution of data and labels using class-specific autoencoders, outperforming existing methods without data storage.

Contribution

It proposes a new generative classification method that learns joint distributions with class-specific autoencoders and importance sampling, improving incremental learning performance.

Findings

Outperforms generative replay and baseline methods on benchmarks.

Effective in rehearsal-free incremental learning scenarios.

Uses variational autoencoders for class-specific modeling.

Abstract

Incrementally training deep neural networks to recognize new classes is a challenging problem. Most existing class-incremental learning methods store data or use generative replay, both of which have drawbacks, while 'rehearsal-free' alternatives such as parameter regularization or bias-correction methods do not consistently achieve high performance. Here, we put forward a new strategy for class-incremental learning: generative classification. Rather than directly learning the conditional distribution p(y|x), our proposal is to learn the joint distribution p(x,y), factorized as p(x|y)p(y), and to perform classification using Bayes' rule. As a proof-of-principle, here we implement this strategy by training a variational autoencoder for each class to be learned and by using importance sampling to estimate the likelihoods p(x|y). This simple approach performs very well on a diverse set of continual learning benchmarks, outperforming generative replay and other existing baselines that do not store data.