OTMatch: Improving Semi-Supervised Learning with Optimal Transport

This article presents OTMatch, a new method for semi-supervised learning that leverages the semantic relationships among classes using an optimal transport loss function. The paper highlights that current algorithms often neglect the inherent relationships within classes. OTMatch has shown superior performance, reducing error rates on CIFAR-10, STL-10, and ImageNet compared to the current state-of-the-art method, FreeMatch. This method demonstrates the effectiveness of harnessing semantic relationships to enhance learning performance in a semi-supervised setting.

Publication date: 26 Oct 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2310.17455