Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Huang, Zhe, Yu, Xiaowei, Zhu, Dajiang, Hughes, Michael C.
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2403.10658
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911799123640320
author Huang, Zhe
Yu, Xiaowei
Zhu, Dajiang
Hughes, Michael C.
author_facet Huang, Zhe
Yu, Xiaowei
Zhu, Dajiang
Hughes, Michael C.
contents Semi-supervised learning (SSL) seeks to enhance task performance by training on both labeled and unlabeled data. Mainstream SSL image classification methods mostly optimize a loss that additively combines a supervised classification objective with a regularization term derived solely from unlabeled data. This formulation neglects the potential for interaction between labeled and unlabeled images. In this paper, we introduce InterLUDE, a new approach to enhance SSL made of two parts that each benefit from labeled-unlabeled interaction. The first part, embedding fusion, interpolates between labeled and unlabeled embeddings to improve representation learning. The second part is a new loss, grounded in the principle of consistency regularization, that aims to minimize discrepancies in the model's predictions between labeled versus unlabeled inputs. Experiments on standard closed-set SSL benchmarks and a medical SSL task with an uncurated unlabeled set show clear benefits to our approach. On the STL-10 dataset with only 40 labels, InterLUDE achieves 3.2% error rate, while the best previous method reports 14.9%.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10658
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning
Huang, Zhe
Yu, Xiaowei
Zhu, Dajiang
Hughes, Michael C.
Computer Vision and Pattern Recognition
Machine Learning
Semi-supervised learning (SSL) seeks to enhance task performance by training on both labeled and unlabeled data. Mainstream SSL image classification methods mostly optimize a loss that additively combines a supervised classification objective with a regularization term derived solely from unlabeled data. This formulation neglects the potential for interaction between labeled and unlabeled images. In this paper, we introduce InterLUDE, a new approach to enhance SSL made of two parts that each benefit from labeled-unlabeled interaction. The first part, embedding fusion, interpolates between labeled and unlabeled embeddings to improve representation learning. The second part is a new loss, grounded in the principle of consistency regularization, that aims to minimize discrepancies in the model's predictions between labeled versus unlabeled inputs. Experiments on standard closed-set SSL benchmarks and a medical SSL task with an uncurated unlabeled set show clear benefits to our approach. On the STL-10 dataset with only 40 labels, InterLUDE achieves 3.2% error rate, while the best previous method reports 14.9%.
title InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2403.10658