Continual learning in low-rank orthogonal subspaces

Arslan Chaudhry*, Naeemullah Khan, Puneet K. Dokania, Philip H.S. Torr

*Corresponding author for this work

Research output: Contribution to conferencePaperpeer-review

34 Scopus citations

Abstract

In continual learning (CL), a learner is faced with a sequence of tasks, arriving one after the other, and the goal is to remember all the tasks once the continual learning experience is finished. The prior art in CL uses episodic memory, parameter regularization or extensible network structures to reduce interference among tasks, but in the end, all the approaches learn different tasks in a joint vector space. We believe this invariably leads to interference among different tasks. We propose to learn tasks in different (low-rank) vector subspaces that are kept orthogonal to each other in order to minimize interference. Further, to keep the gradients of different tasks coming from these subspaces orthogonal to each other, we learn isometric mappings by posing network training as an optimization problem over the Stiefel manifold. To the best of our understanding, we report, for the first time, strong results over experience-replay baseline with and without memory on standard classification benchmarks in continual learning.

Original languageEnglish (US)
StatePublished - 2020
Event34th Conference on Neural Information Processing Systems, NeurIPS 2020 - Virtual, Online
Duration: Dec 6 2020Dec 12 2020

Conference

Conference34th Conference on Neural Information Processing Systems, NeurIPS 2020
CityVirtual, Online
Period12/6/2012/12/20

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Fingerprint

Dive into the research topics of 'Continual learning in low-rank orthogonal subspaces'. Together they form a unique fingerprint.

Cite this