Don't forget, there is more than forgetting: new metrics for Continual Learning

TL;DR

This paper introduces new, comprehensive metrics for evaluating continual learning algorithms, addressing existing gaps by considering accuracy, transfer, memory, and efficiency, and proposes a unified score for better ranking.

Contribution

It proposes a set of implementation-independent metrics for continual learning evaluation and a method to fuse them into a single ranking score.

Findings

The new metrics provide a more holistic evaluation of continual learning algorithms.

The unified score effectively ranks different strategies on the iCIFAR-100 benchmark.

Evaluation shows the metrics capture diverse aspects of continual learning performance.

Abstract

Continual learning consists of algorithms that learn from a stream of data/tasks continuously and adaptively thought time, enabling the incremental development of ever more complex knowledge and skills. The lack of consensus in evaluating continual learning algorithms and the almost exclusive focus on forgetting motivate us to propose a more comprehensive set of implementation independent metrics accounting for several factors we believe have practical implications worth considering in the deployment of real AI systems that learn continually: accuracy or performance over time, backward and forward knowledge transfer, memory overhead as well as computational efficiency. Drawing inspiration from the standard Multi-Attribute Value Theory (MAVT) we further propose to fuse these metrics into a single score for ranking purposes and we evaluate our proposal with five continual learning strategies on the iCIFAR-100 continual learning benchmark.