Amphibian: A Meta-Learner for Rehearsal-Free Fast Online Continual Learning

Gobinda Saha, Kaushik Roy

Primary Area: transfer learning, meta learning, and lifelong learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Continual learning, Meta learning, Online learning, Deep learning Algorithm

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: Our proposed methods- Amphibian meta-learns how to scale the direction of gradient descent without any rehearsal to enable fast online continual learning.

Abstract: Online continual learning is challenging as it requires fast adaptation over a stream of data in a non-stationary environment without forgetting the knowledge acquired in the past. To address this challenge, in this paper, we introduce Amphibian - a gradient-based meta-learner that learns to scale the direction of gradient descent to achieve the desired balance between fast learning and continual learning. For this purpose, using only the current batch of data, Amphibian minimizes a meta-objective that encourages alignments of gradients among given data samples along selected basis directions in the gradient space. From this objective, it learns a diagonal scale matrix in each layer that accumulates the history of such gradient alignments. Using these scale matrices Amphibian updates the model online only in the directions having positive cumulative gradient alignments among the data observed for far. With evaluation on standard continual image classification benchmarks, we show that such meta-learned scaled gradient descent in Amphibian achieves state-of-the-art accuracy in online continual learning while enabling fast learning with less data and few-shot knowledge transfer to new tasks. Finally, with loss landscape visualizations, we show such gradient updates incur minimum loss to the old task enabling fast continual learning in Amphibian.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 6637