Subspace Learning for Effective Meta-Learning

Weisen Jiang, James Kwok, Yu Zhang
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:10177-10194, 2022.

Abstract

Meta-learning aims to extract meta-knowledge from historical tasks to accelerate learning on new tasks. Typical meta-learning algorithms like MAML learn a globally-shared meta-model for all tasks. However, when the task environments are complex, task model parameters are diverse and a common meta-model is insufficient to capture all the meta-knowledge. To address this challenge, in this paper, task model parameters are structured into multiple subspaces, and each subspace represents one type of meta-knowledge. We propose an algorithm to learn the meta-parameters (\ie, subspace bases). We theoretically study the generalization properties of the learned subspaces. Experiments on regression and classification meta-learning datasets verify the effectiveness of the proposed algorithm.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-jiang22b, title = {Subspace Learning for Effective Meta-Learning}, author = {Jiang, Weisen and Kwok, James and Zhang, Yu}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {10177--10194}, 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/jiang22b/jiang22b.pdf}, url = {https://proceedings.mlr.press/v162/jiang22b.html}, abstract = {Meta-learning aims to extract meta-knowledge from historical tasks to accelerate learning on new tasks. Typical meta-learning algorithms like MAML learn a globally-shared meta-model for all tasks. However, when the task environments are complex, task model parameters are diverse and a common meta-model is insufficient to capture all the meta-knowledge. To address this challenge, in this paper, task model parameters are structured into multiple subspaces, and each subspace represents one type of meta-knowledge. We propose an algorithm to learn the meta-parameters (\ie, subspace bases). We theoretically study the generalization properties of the learned subspaces. Experiments on regression and classification meta-learning datasets verify the effectiveness of the proposed algorithm.} }
Endnote
%0 Conference Paper %T Subspace Learning for Effective Meta-Learning %A Weisen Jiang %A James Kwok %A Yu Zhang %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-jiang22b %I PMLR %P 10177--10194 %U https://proceedings.mlr.press/v162/jiang22b.html %V 162 %X Meta-learning aims to extract meta-knowledge from historical tasks to accelerate learning on new tasks. Typical meta-learning algorithms like MAML learn a globally-shared meta-model for all tasks. However, when the task environments are complex, task model parameters are diverse and a common meta-model is insufficient to capture all the meta-knowledge. To address this challenge, in this paper, task model parameters are structured into multiple subspaces, and each subspace represents one type of meta-knowledge. We propose an algorithm to learn the meta-parameters (\ie, subspace bases). We theoretically study the generalization properties of the learned subspaces. Experiments on regression and classification meta-learning datasets verify the effectiveness of the proposed algorithm.
APA
Jiang, W., Kwok, J. & Zhang, Y.. (2022). Subspace Learning for Effective Meta-Learning. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:10177-10194 Available from https://proceedings.mlr.press/v162/jiang22b.html.

Related Material