Online Continual Learning through Mutual Information Maximization

Yiduo Guo, Bing Liu, Dongyan Zhao
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:8109-8126, 2022.

Abstract

This paper proposed a new online continual learning approach called OCM based on mutual information (MI) maximization. It achieves two objectives that are critical in dealing with catastrophic forgetting (CF). (1) It reduces feature bias caused by cross entropy (CE) as CE learns only discriminative features for each task, but these features may not be discriminative for another task. To learn a new task well, the network parameters learned before have to be modified, which causes CF. The new approach encourages the learning of each task to make use of the full features of the task training data. (2) It encourages preservation of the previously learned knowledge when training a new batch of incrementally arriving data. Empirical evaluation shows that OCM substantially outperforms the latest online CL baselines. For example, for CIFAR10, OCM improves the accuracy of the best baseline by 13.1% from 64.1% (baseline) to 77.2% (OCM).The code is publicly available at https://github.com/gydpku/OCM.

Cite this Paper


BibTeX
@InProceedings{pmlr-v162-guo22g, title = {Online Continual Learning through Mutual Information Maximization}, author = {Guo, Yiduo and Liu, Bing and Zhao, Dongyan}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {8109--8126}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/guo22g/guo22g.pdf}, url = {https://proceedings.mlr.press/v162/guo22g.html}, abstract = {This paper proposed a new online continual learning approach called OCM based on mutual information (MI) maximization. It achieves two objectives that are critical in dealing with catastrophic forgetting (CF). (1) It reduces feature bias caused by cross entropy (CE) as CE learns only discriminative features for each task, but these features may not be discriminative for another task. To learn a new task well, the network parameters learned before have to be modified, which causes CF. The new approach encourages the learning of each task to make use of the full features of the task training data. (2) It encourages preservation of the previously learned knowledge when training a new batch of incrementally arriving data. Empirical evaluation shows that OCM substantially outperforms the latest online CL baselines. For example, for CIFAR10, OCM improves the accuracy of the best baseline by 13.1% from 64.1% (baseline) to 77.2% (OCM).The code is publicly available at https://github.com/gydpku/OCM.} }
Endnote
%0 Conference Paper %T Online Continual Learning through Mutual Information Maximization %A Yiduo Guo %A Bing Liu %A Dongyan Zhao %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-guo22g %I PMLR %P 8109--8126 %U https://proceedings.mlr.press/v162/guo22g.html %V 162 %X This paper proposed a new online continual learning approach called OCM based on mutual information (MI) maximization. It achieves two objectives that are critical in dealing with catastrophic forgetting (CF). (1) It reduces feature bias caused by cross entropy (CE) as CE learns only discriminative features for each task, but these features may not be discriminative for another task. To learn a new task well, the network parameters learned before have to be modified, which causes CF. The new approach encourages the learning of each task to make use of the full features of the task training data. (2) It encourages preservation of the previously learned knowledge when training a new batch of incrementally arriving data. Empirical evaluation shows that OCM substantially outperforms the latest online CL baselines. For example, for CIFAR10, OCM improves the accuracy of the best baseline by 13.1% from 64.1% (baseline) to 77.2% (OCM).The code is publicly available at https://github.com/gydpku/OCM.
APA
Guo, Y., Liu, B. & Zhao, D.. (2022). Online Continual Learning through Mutual Information Maximization. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:8109-8126 Available from https://proceedings.mlr.press/v162/guo22g.html.

Related Material