Subspace Distillation for Continual Learning

TL;DR

This paper introduces a subspace-based knowledge distillation method for continual learning that models the data manifold to reduce forgetting, demonstrating improved performance on multiple datasets and compatibility with existing approaches.

Contribution

The paper proposes a novel subspace distillation technique that models the data manifold's structure to mitigate catastrophic forgetting in continual learning.

Findings

Outperforms existing methods on Pascal VOC and Tiny-Imagenet

Provides robustness to noise in continual learning scenarios

Can be integrated with other learning approaches for enhanced performance

Abstract

An ultimate objective in continual learning is to preserve knowledge learned in preceding tasks while learning new tasks. To mitigate forgetting prior knowledge, we propose a novel knowledge distillation technique that takes into the account the manifold structure of the latent/output space of a neural network in learning novel tasks. To achieve this, we propose to approximate the data manifold up-to its first order, hence benefiting from linear subspaces to model the structure and maintain the knowledge of a neural network while learning novel concepts. We demonstrate that the modeling with subspaces provides several intriguing properties, including robustness to noise and therefore effective for mitigating Catastrophic Forgetting in continual learning. We also discuss and show how our proposed method can be adopted to address both classification and segmentation problems. Empirically, we observe that our proposed method outperforms various continual learning methods on several challenging datasets including Pascal VOC, and Tiny-Imagenet. Furthermore, we show how the proposed method can be seamlessly combined with existing learning approaches to improve their performances. The codes of this article will be available at https://github.com/csiro-robotics/SDCL.