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

TL;DR

This paper introduces FineSSL, a novel semi-supervised learning method that fine-tunes foundation models by addressing inherent biases, achieving state-of-the-art results, and significantly reducing training costs.

Contribution

The paper proposes a new SSL approach, FineSSL, which effectively mitigates biases in foundation models through balanced margin softmax and label smoothing, improving performance and efficiency.

Findings

Sets new state-of-the-art on multiple benchmarks.

Reduces training cost by over six times.

Easily integrates with various SSL algorithms.

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.