Insights from the Future for Continual Learning

TL;DR

This paper introduces prescient continual learning, a new setting where models leverage future class information without training samples, using a Ghost Model to improve class representation and mitigate catastrophic forgetting.

Contribution

The work proposes a novel prescient continual learning setting and a Ghost Model that incorporates future class insights to enhance learning without future training data.

Findings

Effective on AwA2 and aP extbar Y datasets

Shows improved class separation in representation space

Demonstrates potential of future class information in continual learning

Abstract

Continual learning aims to learn tasks sequentially, with (often severe) constraints on the storage of old learning samples, without suffering from catastrophic forgetting. In this work, we propose prescient continual learning, a novel experimental setting, to incorporate existing information about the classes, prior to any training data. Usually, each task in a traditional continual learning setting evaluates the model on present and past classes, the latter with a limited number of training samples. Our setting adds future classes, with no training samples at all. We introduce Ghost Model, a representation-learning-based model for continual learning using ideas from zero-shot learning. A generative model of the representation space in concert with a careful adjustment of the losses allows us to exploit insights from future classes to constraint the spatial arrangement of the past and current classes. Quantitative results on the AwA2 and aP\&Y datasets and detailed visualizations showcase the interest of this new setting and the method we propose to address it.