Infinite Mixture Prototypes for Few-shot Learning

Kelsey Allen, Evan Shelhamer, Hanul Shin, Joshua Tenenbaum
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:232-241, 2019.

Abstract

We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Infinite mixture prototypes combine deep representation learning with Bayesian nonparametrics, representing each class by a set of clusters, unlike existing prototypical methods that represent each class by a single cluster. By inferring the number of clusters, infinite mixture prototypes interpolate between nearest neighbor and prototypical representations in a learned feature space, which improves accuracy and robustness in the few-shot regime. We show the importance of adaptive capacity for capturing complex data distributions such as super-classes (like alphabets in character recognition), with 10-25% absolute accuracy improvements over prototypical networks, while still maintaining or improving accuracy on standard few-shot learning benchmarks. By clustering labeled and unlabeled data with the same rule, infinite mixture prototypes achieve state-of-the-art semi-supervised accuracy, and can perform purely unsupervised clustering, unlike existing fully- and semi-supervised prototypical methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-allen19b, title = {Infinite Mixture Prototypes for Few-shot Learning}, author = {Allen, Kelsey and Shelhamer, Evan and Shin, Hanul and Tenenbaum, Joshua}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {232--241}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/allen19b/allen19b.pdf}, url = {https://proceedings.mlr.press/v97/allen19b.html}, abstract = {We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Infinite mixture prototypes combine deep representation learning with Bayesian nonparametrics, representing each class by a set of clusters, unlike existing prototypical methods that represent each class by a single cluster. By inferring the number of clusters, infinite mixture prototypes interpolate between nearest neighbor and prototypical representations in a learned feature space, which improves accuracy and robustness in the few-shot regime. We show the importance of adaptive capacity for capturing complex data distributions such as super-classes (like alphabets in character recognition), with 10-25% absolute accuracy improvements over prototypical networks, while still maintaining or improving accuracy on standard few-shot learning benchmarks. By clustering labeled and unlabeled data with the same rule, infinite mixture prototypes achieve state-of-the-art semi-supervised accuracy, and can perform purely unsupervised clustering, unlike existing fully- and semi-supervised prototypical methods.} }
Endnote
%0 Conference Paper %T Infinite Mixture Prototypes for Few-shot Learning %A Kelsey Allen %A Evan Shelhamer %A Hanul Shin %A Joshua Tenenbaum %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-allen19b %I PMLR %P 232--241 %U https://proceedings.mlr.press/v97/allen19b.html %V 97 %X We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Infinite mixture prototypes combine deep representation learning with Bayesian nonparametrics, representing each class by a set of clusters, unlike existing prototypical methods that represent each class by a single cluster. By inferring the number of clusters, infinite mixture prototypes interpolate between nearest neighbor and prototypical representations in a learned feature space, which improves accuracy and robustness in the few-shot regime. We show the importance of adaptive capacity for capturing complex data distributions such as super-classes (like alphabets in character recognition), with 10-25% absolute accuracy improvements over prototypical networks, while still maintaining or improving accuracy on standard few-shot learning benchmarks. By clustering labeled and unlabeled data with the same rule, infinite mixture prototypes achieve state-of-the-art semi-supervised accuracy, and can perform purely unsupervised clustering, unlike existing fully- and semi-supervised prototypical methods.
APA
Allen, K., Shelhamer, E., Shin, H. & Tenenbaum, J.. (2019). Infinite Mixture Prototypes for Few-shot Learning. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:232-241 Available from https://proceedings.mlr.press/v97/allen19b.html.

Related Material