Data Augmentation for Meta-Learning

Renkun Ni, Micah Goldblum, Amr Sharaf, Kezhi Kong, Tom Goldstein
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:8152-8161, 2021.

Abstract

Conventional image classifiers are trained by randomly sampling mini-batches of images. To achieve state-of-the-art performance, practitioners use sophisticated data augmentation schemes to expand the amount of training data available for sampling. In contrast, meta-learning algorithms sample support data, query data, and tasks on each training step. In this complex sampling scenario, data augmentation can be used not only to expand the number of images available per class, but also to generate entirely new classes/tasks. We systematically dissect the meta-learning pipeline and investigate the distinct ways in which data augmentation can be integrated at both the image and class levels. Our proposed meta-specific data augmentation significantly improves the performance of meta-learners on few-shot classification benchmarks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-ni21a, title = {Data Augmentation for Meta-Learning}, author = {Ni, Renkun and Goldblum, Micah and Sharaf, Amr and Kong, Kezhi and Goldstein, Tom}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {8152--8161}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/ni21a/ni21a.pdf}, url = {https://proceedings.mlr.press/v139/ni21a.html}, abstract = {Conventional image classifiers are trained by randomly sampling mini-batches of images. To achieve state-of-the-art performance, practitioners use sophisticated data augmentation schemes to expand the amount of training data available for sampling. In contrast, meta-learning algorithms sample support data, query data, and tasks on each training step. In this complex sampling scenario, data augmentation can be used not only to expand the number of images available per class, but also to generate entirely new classes/tasks. We systematically dissect the meta-learning pipeline and investigate the distinct ways in which data augmentation can be integrated at both the image and class levels. Our proposed meta-specific data augmentation significantly improves the performance of meta-learners on few-shot classification benchmarks.} }
Endnote
%0 Conference Paper %T Data Augmentation for Meta-Learning %A Renkun Ni %A Micah Goldblum %A Amr Sharaf %A Kezhi Kong %A Tom Goldstein %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-ni21a %I PMLR %P 8152--8161 %U https://proceedings.mlr.press/v139/ni21a.html %V 139 %X Conventional image classifiers are trained by randomly sampling mini-batches of images. To achieve state-of-the-art performance, practitioners use sophisticated data augmentation schemes to expand the amount of training data available for sampling. In contrast, meta-learning algorithms sample support data, query data, and tasks on each training step. In this complex sampling scenario, data augmentation can be used not only to expand the number of images available per class, but also to generate entirely new classes/tasks. We systematically dissect the meta-learning pipeline and investigate the distinct ways in which data augmentation can be integrated at both the image and class levels. Our proposed meta-specific data augmentation significantly improves the performance of meta-learners on few-shot classification benchmarks.
APA
Ni, R., Goldblum, M., Sharaf, A., Kong, K. & Goldstein, T.. (2021). Data Augmentation for Meta-Learning. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:8152-8161 Available from https://proceedings.mlr.press/v139/ni21a.html.

Related Material