Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning

Kai Gan, Tong Wei
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:14453-14470, 2024.

Abstract

Semi-supervised learning (SSL) has witnessed remarkable progress, resulting in the emergence of numerous method variations. However, practitioners often encounter challenges when attempting to deploy these methods due to their subpar performance. In this paper, we present a novel SSL approach named FineSSL that significantly addresses this limitation by adapting pre-trained foundation models. We identify the aggregated biases and cognitive deviation problems inherent in foundation models, and propose a simple yet effective solution by imposing balanced margin softmax and decoupled label smoothing. Through extensive experiments, we demonstrate that FineSSL sets a new state of the art for SSL on multiple benchmark datasets, reduces the training cost by over six times, and can seamlessly integrate various fine-tuning and modern SSL algorithms. The source code is available at https://github.com/Gank0078/FineSSL.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-gan24a, title = {Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning}, author = {Gan, Kai and Wei, Tong}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {14453--14470}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/gan24a/gan24a.pdf}, url = {https://proceedings.mlr.press/v235/gan24a.html}, abstract = {Semi-supervised learning (SSL) has witnessed remarkable progress, resulting in the emergence of numerous method variations. However, practitioners often encounter challenges when attempting to deploy these methods due to their subpar performance. In this paper, we present a novel SSL approach named FineSSL that significantly addresses this limitation by adapting pre-trained foundation models. We identify the aggregated biases and cognitive deviation problems inherent in foundation models, and propose a simple yet effective solution by imposing balanced margin softmax and decoupled label smoothing. Through extensive experiments, we demonstrate that FineSSL sets a new state of the art for SSL on multiple benchmark datasets, reduces the training cost by over six times, and can seamlessly integrate various fine-tuning and modern SSL algorithms. The source code is available at https://github.com/Gank0078/FineSSL.} }
Endnote
%0 Conference Paper %T Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning %A Kai Gan %A Tong Wei %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-gan24a %I PMLR %P 14453--14470 %U https://proceedings.mlr.press/v235/gan24a.html %V 235 %X Semi-supervised learning (SSL) has witnessed remarkable progress, resulting in the emergence of numerous method variations. However, practitioners often encounter challenges when attempting to deploy these methods due to their subpar performance. In this paper, we present a novel SSL approach named FineSSL that significantly addresses this limitation by adapting pre-trained foundation models. We identify the aggregated biases and cognitive deviation problems inherent in foundation models, and propose a simple yet effective solution by imposing balanced margin softmax and decoupled label smoothing. Through extensive experiments, we demonstrate that FineSSL sets a new state of the art for SSL on multiple benchmark datasets, reduces the training cost by over six times, and can seamlessly integrate various fine-tuning and modern SSL algorithms. The source code is available at https://github.com/Gank0078/FineSSL.
APA
Gan, K. & Wei, T.. (2024). Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:14453-14470 Available from https://proceedings.mlr.press/v235/gan24a.html.

Related Material