Automating Continual Learning

Kazuki Irie, Róbert Csordás, Jürgen Schmidhuber

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, in-context learning, meta-learning, self-referential learning, linear Transformers

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

TL;DR: We train self-referential neural networks that learn continual learning algorithms

Abstract:

General-purpose learning systems should improve themselves in open-ended fashion in ever-changing environments. Conventional learning algorithms for neural networks, however, suffer from the so-called catastrophic forgetting (CF) problem---previously acquired skills are forgotten when a new task is learned. Developing continual learning algorithms to address CF remains an open research question.
Instead of hand-crafting such algorithms, our new Automated Continual Learning (ACL) trains self-referential neural networks to meta-learn their own in-context continual (meta-)learning algorithms. ACL encodes all desiderata---good performance on both old and new tasks---into its learning objectives. We demonstrate the effectiveness and promise of ACL on multiple few-shot and standard image classification tasks adopted for continual learning: Mini-ImageNet, Omniglot, FC100, MNIST-families, and CIFAR-10.

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: 9084