Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels

Min-Kook Suh, Seung-Woo Seo
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:32770-32782, 2023.

Abstract

Although contrastive learning methods have shown prevailing performance on a variety of representation learning tasks, they encounter difficulty when the training dataset is long-tailed. Many researchers have combined contrastive learning and a logit adjustment technique to address this problem, but the combinations are done ad-hoc and a theoretical background has not yet been provided. The goal of this paper is to provide the background and further improve the performance. First, we show that the fundamental reason contrastive learning methods struggle with long-tailed tasks is that they try to maximize the mutual information between latent features and input data. As ground-truth labels are not considered in the maximization, they are not able to address imbalances between classes. Rather, we interpret the long-tailed recognition task as a mutual information maximization between latent features and ground-truth labels. This approach integrates contrastive learning and logit adjustment seamlessly to derive a loss function that shows state-of-the-art performance on long-tailed recognition benchmarks. It also demonstrates its efficacy in image segmentation tasks, verifying its versatility beyond image classification. Code is available at https://github.com/bluecdm/Long-tailed-recognition.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-suh23a, title = {Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels}, author = {Suh, Min-Kook and Seo, Seung-Woo}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {32770--32782}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/suh23a/suh23a.pdf}, url = {https://proceedings.mlr.press/v202/suh23a.html}, abstract = {Although contrastive learning methods have shown prevailing performance on a variety of representation learning tasks, they encounter difficulty when the training dataset is long-tailed. Many researchers have combined contrastive learning and a logit adjustment technique to address this problem, but the combinations are done ad-hoc and a theoretical background has not yet been provided. The goal of this paper is to provide the background and further improve the performance. First, we show that the fundamental reason contrastive learning methods struggle with long-tailed tasks is that they try to maximize the mutual information between latent features and input data. As ground-truth labels are not considered in the maximization, they are not able to address imbalances between classes. Rather, we interpret the long-tailed recognition task as a mutual information maximization between latent features and ground-truth labels. This approach integrates contrastive learning and logit adjustment seamlessly to derive a loss function that shows state-of-the-art performance on long-tailed recognition benchmarks. It also demonstrates its efficacy in image segmentation tasks, verifying its versatility beyond image classification. Code is available at https://github.com/bluecdm/Long-tailed-recognition.} }
Endnote
%0 Conference Paper %T Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels %A Min-Kook Suh %A Seung-Woo Seo %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-suh23a %I PMLR %P 32770--32782 %U https://proceedings.mlr.press/v202/suh23a.html %V 202 %X Although contrastive learning methods have shown prevailing performance on a variety of representation learning tasks, they encounter difficulty when the training dataset is long-tailed. Many researchers have combined contrastive learning and a logit adjustment technique to address this problem, but the combinations are done ad-hoc and a theoretical background has not yet been provided. The goal of this paper is to provide the background and further improve the performance. First, we show that the fundamental reason contrastive learning methods struggle with long-tailed tasks is that they try to maximize the mutual information between latent features and input data. As ground-truth labels are not considered in the maximization, they are not able to address imbalances between classes. Rather, we interpret the long-tailed recognition task as a mutual information maximization between latent features and ground-truth labels. This approach integrates contrastive learning and logit adjustment seamlessly to derive a loss function that shows state-of-the-art performance on long-tailed recognition benchmarks. It also demonstrates its efficacy in image segmentation tasks, verifying its versatility beyond image classification. Code is available at https://github.com/bluecdm/Long-tailed-recognition.
APA
Suh, M. & Seo, S.. (2023). Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:32770-32782 Available from https://proceedings.mlr.press/v202/suh23a.html.

Related Material