Retrospective Feature Estimation for Continual Learning

TL;DR

This paper introduces Retrospective Feature Estimation (RFE), a novel continual learning method that reverses feature changes to mitigate catastrophic forgetting, showing promising results on standard benchmarks.

Contribution

It proposes RFE, a new approach that aligns current features to past task features using retrospector modules, offering an alternative to existing CL methods.

Findings

RFE outperforms some existing CL methods on CIFAR benchmarks.

Retrospective feature alignment reduces catastrophic forgetting effectively.

The approach demonstrates potential for further research in CL mechanisms.

Abstract

The intrinsic capability to continuously learn a changing data stream is a desideratum of deep neural networks (DNNs). However, current DNNs suffer from catastrophic forgetting, which interferes with remembering past knowledge. To mitigate this issue, existing Continual Learning (CL) approaches often retain exemplars for replay, regularize learning, or allocate dedicated capacity for new tasks. This paper investigates an unexplored direction for CL called Retrospective Feature Estimation (RFE). RFE learns to reverse feature changes by aligning the features from the current trained DNN backward to the feature space of the old task, where performing predictions is easier. This retrospective process utilizes a chain of small feature mapping networks called retrospector modules. Empirical experiments on several CL benchmarks, including CIFAR10, CIFAR100, and Tiny ImageNet, demonstrate the effectiveness and potential of this novel CL direction compared to existing representative CL methods, motivating further research into retrospective mechanisms as a principled alternative for mitigating catastrophic forgetting in CL. Code is available at: https://github.com/mail-research/retrospective-feature-estimation.