Energy-Based Models for Continual Learning

Shuang Li, Yilun Du, Gido van de Ven, Igor Mordatch
Proceedings of The 1st Conference on Lifelong Learning Agents, PMLR 199:1-22, 2022.

Abstract

We motivate Energy-Based Models (EBMs) as a promising model class for continual learning problems. Instead of tackling continual learning via the use of external memory, growing models, or regularization, EBMs change the underlying training objective to cause less interference with previously learned information. Our proposed version of EBMs for continual learning is simple, efficient, and outperforms baseline methods by a large margin on several benchmarks. Moreover, our proposed contrastive divergence-based training objective can be combined with other continual learning methods, resulting in substantial boosts in their performance. We further show that EBMs are adaptable to a more general continual learning setting where the data distribution changes without the notion of explicitly delineated tasks. These observations point towards EBMs as a useful building block for future continual learning methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v199-li22a, title = {Energy-Based Models for Continual Learning}, author = {Li, Shuang and Du, Yilun and van de Ven, Gido and Mordatch, Igor}, booktitle = {Proceedings of The 1st Conference on Lifelong Learning Agents}, pages = {1--22}, year = {2022}, editor = {Chandar, Sarath and Pascanu, Razvan and Precup, Doina}, volume = {199}, series = {Proceedings of Machine Learning Research}, month = {22--24 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v199/li22a/li22a.pdf}, url = {https://proceedings.mlr.press/v199/li22a.html}, abstract = {We motivate Energy-Based Models (EBMs) as a promising model class for continual learning problems. Instead of tackling continual learning via the use of external memory, growing models, or regularization, EBMs change the underlying training objective to cause less interference with previously learned information. Our proposed version of EBMs for continual learning is simple, efficient, and outperforms baseline methods by a large margin on several benchmarks. Moreover, our proposed contrastive divergence-based training objective can be combined with other continual learning methods, resulting in substantial boosts in their performance. We further show that EBMs are adaptable to a more general continual learning setting where the data distribution changes without the notion of explicitly delineated tasks. These observations point towards EBMs as a useful building block for future continual learning methods.} }
Endnote
%0 Conference Paper %T Energy-Based Models for Continual Learning %A Shuang Li %A Yilun Du %A Gido van de Ven %A Igor Mordatch %B Proceedings of The 1st Conference on Lifelong Learning Agents %C Proceedings of Machine Learning Research %D 2022 %E Sarath Chandar %E Razvan Pascanu %E Doina Precup %F pmlr-v199-li22a %I PMLR %P 1--22 %U https://proceedings.mlr.press/v199/li22a.html %V 199 %X We motivate Energy-Based Models (EBMs) as a promising model class for continual learning problems. Instead of tackling continual learning via the use of external memory, growing models, or regularization, EBMs change the underlying training objective to cause less interference with previously learned information. Our proposed version of EBMs for continual learning is simple, efficient, and outperforms baseline methods by a large margin on several benchmarks. Moreover, our proposed contrastive divergence-based training objective can be combined with other continual learning methods, resulting in substantial boosts in their performance. We further show that EBMs are adaptable to a more general continual learning setting where the data distribution changes without the notion of explicitly delineated tasks. These observations point towards EBMs as a useful building block for future continual learning methods.
APA
Li, S., Du, Y., van de Ven, G. & Mordatch, I.. (2022). Energy-Based Models for Continual Learning. Proceedings of The 1st Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 199:1-22 Available from https://proceedings.mlr.press/v199/li22a.html.

Related Material