Online Continual Learning for Embedded Devices

Tyler L. Hayes, Christopher Kanan
Proceedings of The 1st Conference on Lifelong Learning Agents, PMLR 199:744-766, 2022.

Abstract

Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets. However, this setting poses unique challenges: embedded devices have limited memory and compute capacity and conventional machine learning models suffer from catastrophic forgetting when updated on non-stationary data streams. While several online continual learning models have been developed, their effectiveness for embedded applications has not been rigorously studied. In this paper, we first identify criteria that online continual learners must meet to effectively perform real-time, on-device learning. We then study the efficacy of several online continual learning methods when used with mobile neural networks. We measure their performance, memory usage, compute requirements, and ability to generalize to out-of-domain inputs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v199-hayes22a, title = {Online Continual Learning for Embedded Devices}, author = {Hayes, Tyler L. and Kanan, Christopher}, booktitle = {Proceedings of The 1st Conference on Lifelong Learning Agents}, pages = {744--766}, 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/hayes22a/hayes22a.pdf}, url = {https://proceedings.mlr.press/v199/hayes22a.html}, abstract = {Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets. However, this setting poses unique challenges: embedded devices have limited memory and compute capacity and conventional machine learning models suffer from catastrophic forgetting when updated on non-stationary data streams. While several online continual learning models have been developed, their effectiveness for embedded applications has not been rigorously studied. In this paper, we first identify criteria that online continual learners must meet to effectively perform real-time, on-device learning. We then study the efficacy of several online continual learning methods when used with mobile neural networks. We measure their performance, memory usage, compute requirements, and ability to generalize to out-of-domain inputs.} }
Endnote
%0 Conference Paper %T Online Continual Learning for Embedded Devices %A Tyler L. Hayes %A Christopher Kanan %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-hayes22a %I PMLR %P 744--766 %U https://proceedings.mlr.press/v199/hayes22a.html %V 199 %X Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets. However, this setting poses unique challenges: embedded devices have limited memory and compute capacity and conventional machine learning models suffer from catastrophic forgetting when updated on non-stationary data streams. While several online continual learning models have been developed, their effectiveness for embedded applications has not been rigorously studied. In this paper, we first identify criteria that online continual learners must meet to effectively perform real-time, on-device learning. We then study the efficacy of several online continual learning methods when used with mobile neural networks. We measure their performance, memory usage, compute requirements, and ability to generalize to out-of-domain inputs.
APA
Hayes, T.L. & Kanan, C.. (2022). Online Continual Learning for Embedded Devices. Proceedings of The 1st Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 199:744-766 Available from https://proceedings.mlr.press/v199/hayes22a.html.

Related Material