Marginal Replay vs Conditional Replay for Continual Learning

TL;DR

This paper introduces 'conditional replay,' a new replay-based continual learning method that generates samples with labels conditioned on class, reducing errors and improving performance compared to marginal replay and EWC on MNIST-based benchmarks.

Contribution

The paper proposes a novel conditional replay approach for continual learning, demonstrating its advantages over marginal replay and existing regularization methods.

Findings

Conditional replay reduces label inference errors.

Conditional replay outperforms marginal replay and EWC on benchmarks.

The method is effective on MNIST and FashionMNIST datasets.

Abstract

We present a new replay-based method of continual classification learning that we term "conditional replay" which generates samples and labels together by sampling from a distribution conditioned on the class. We compare conditional replay to another replay-based continual learning paradigm (which we term "marginal replay") that generates samples independently of their class and assigns labels in a separate step. The main improvement in conditional replay is that labels for generated samples need not be inferred, which reduces the margin for error in complex continual classification learning tasks. We demonstrate the effectiveness of this approach using novel and standard benchmarks constructed from MNIST and FashionMNIST data, and compare to the regularization-based \textit{elastic weight consolidation} (EWC) method.