2024年3月18日，计算机类top期刊《Information Sciences》在线发表了云南师范大学信息学院户战选在《Information Sciences》发表最新研究成果《Neural collapse inspired semi-supervised learning with fixed classifier》。云南师范大学信息学院户战选为第一作者，物理与电子信息学院邰永航为通讯作者。

https://www.sciencedirect.com/science/article/pii/S0020025524003827#fg0020 Abstract

Pseudo-labeling-based approaches are gaining prominence in Semi-Supervised Learning (SSL). Recent studies have identified that the key bottleneck in this methodology is addressing insufficient, incorrect, and imbalanced pseudo-labels. In this paper, we argue that the intrinsic problem behind this bottleneck is classifier bias, i.e., the classifier's prototypes suffer from poor uniformity. Further, inspired by neural collapse that reveals an optimal structure under supervised training scenarios, we address this classifier bias by utilizing an offline simplex Equiangular Tight Frame (ETF) classifier with maximally and equally separated prototypes. During the training phase, we maintain the prototypes of the classifier as fixed and concentrate on refining the feature encoder. Specifically, we integrate a straightforward clustering-based pseudo-labeling strategy with information maximization for feature learning. In practice, the fixed ETF classifier prevents the model from falling into a detrimental cycle, where a biased classifier exacerbates misaligned features, further perpetuating this bias. Furthermore, the clustering-based pseudo-labeling strategy reduces the dependency on complex threshold-adjusting mechanisms and effectively navigates the quantity-quality trade-off that plagues existing SSL methods. Leveraging these methodologies, we develop a simple yet powerful approach, termed ETF-SSL. Extensive experiments across Image, Text, and Audio datasets demonstrate that ETF-SSL can achieve competitive or superior performance compared to existing approaches. This success highlights the benefits of using a fixed ETF classifier in SSL and points to promising directions for future research in this area. The code is available at: https://github.com/yichenwang231/ETFSSL.

拓展阅读：

https://cic.ynnu.edu.cn/info/1012/2321.htm

个人简介

户战选，男，汉族，1991年1月出生，河南濮阳人，博士，副教授，硕士生导师。2021年6月毕业于西北工业大学光电与智能研究院（导师：聂飞平），获工学博士学位。主要研究方向为标注有限场景下的计算机视觉、数据挖掘和时间序列分析问题。主持国家自然科学基金青年项目1项。近年，在《Information Fusion》、《Information Sciences》、《IEEE Transactions on lmage Processing》、《IEEE Transactions on Cybernetics》、KDD、AAAI等国际知名会议和期刊上发表学术论文20余篇。长期担任IEEE-TNNLS、IEEE-TCSVT、Neurocomputing等期刊审稿人。

研究方向

人工智能、计算机视觉、时间序列分析

工作经历

2021.7-2024.1  西安邮电大学     计算机学院 (副教授)

2024.2-至今  云南师范大学     信息学院（副教授）

科研项目

1、2023-2025 基于代表性子集标注的大规模图像表征学习方法研究，国家自然科学基金青年项目 主持

发表论文

1. Qingsong Hu, Huafeng Li, Zhanxuan Hu*, Feiping Nie. Diverse semantic information fusion for unsupervised person re-identification[J]. Information Fusion, 2024: 102319.

2. Zhanxuan Hu, Yichen Wang, Hailong Ning, Danyang Wu, Feiping Nie. " Mutual-Taught Deep Clustering ". Knowledge-Based Systems, 2023.

3. Huafeng Li, Minghui Liu, Zhanxuan Hu*, Zhengtao Yu. " Intermediary-guided Bidirectional Spatial Temporal Aggregation Network for Video-based Visible-Infrared Person Re-Identification". IEEE Transactions on Circuits and Systems for Video Technology, 2023.

4. Danyng Wu, Han Wang, Zhanxuan Hu*, Feiping Nie " Improved deep metric learning with local neighborhood component analysis". Information Sciences, 2022.

5. Zhanxuan Hu*, Danyang Wu*, Feiping Nie, and Rong Wang " Generalization Bottleneck in Deep Metric Learning". Information Sciences, 2021.

6. Zhanxuan Hu, Feiping Nie, Rong Wang*, and Xuelong Li, "Multi-view spectral clustering via integrating nonnegative embedding and spectral embedding," Information Fusion, vol.55, pp. 251-259, 2020.

联系方式:

邮件：zhanxuanhu@gmail.com