Dynamic Weight-Enhanced Contrastive Learning for Partially View-Aligned Clustering
IEEE Transactions on Circuits and Systems for Video Technology · 2026-08 · 36 (8) · 12426-12438
DOI: 10.1109/tcsvt.2026.3679096 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-08. Verfasst von Qiuyu Chen und 5 weiteren Autor:innen. Die Arbeit nennt 64 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project.
Abstract (Originalsprache)
Deep contrastive multi-view clustering has attracted considerable attention for its ability to extract shared representations across heterogeneous views. Despite its effectiveness, existing approaches usually assume full cross-view alignment, an assumption often violated in practice due to sensor noise, transmission loss, or inconsistent data augmentation. This partial view alignment substantially degrades performance, while current solutions relying on fixed pair selection or uniform weighting fail to resolve the semantic ambiguity of unaligned samples. Consequently, latent cross-view semantic relationships remain underutilized, while erroneous positive pairs unduly dominate the contrastive objective, compromising robustness under low alignment rates or noisy conditions. To address these challenges, we propose Dynamic Weight-Enhanced InfoNCE (DW-InfoNCE), a contrastive learning framework designed for partially view-aligned clustering. The framework introduces a cross-view positive pair mining strategy that iteratively identifies reliable pairs from unaligned samples through quantile-adaptive thresholds and uniqueness constraints, enabling recovery of latent semantic structures obscured by incomplete alignment. In addition, an adaptive faulty pair suppression mechanism is developed to identify semantically inconsistent pairs via temporal loss smoothing and global statistics, while attenuating their impact with Gaussian-based adaptive weights. Extensive experiments on five benchmarks demonstrate DW-InfoNCE consistently outperforms state-of-the-art methods across alignment rates, with significant gains under severe misalignment. The code of DW-InfoNCE can be publicly downloaded at https://github.com/qiuyu99627/DWInfoNCE.git
Förderung
- National Natural Science Foundation of China
- Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project
Zitieren
APA
Chen, Q., Yang, X., Dong, Z., Qu, Q., Zhu, E., & Liu, X. (2026). Dynamic Weight-Enhanced Contrastive Learning for Partially View-Aligned Clustering. IEEE Transactions on Circuits and Systems for Video Technology, 36(8), 12426-12438. https://doi.org/10.1109/tcsvt.2026.3679096
Vancouver
Chen Q, Yang X, Dong Z, Qu Q, Zhu E, Liu X. Dynamic Weight-Enhanced Contrastive Learning for Partially View-Aligned Clustering. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(8):12426-12438. doi:10.1109/tcsvt.2026.3679096
BibTeX
@article{Chen2026,
author = {Chen, Qiuyu and Yang, Xihong and Dong, Zhibing and Qu, Qian and Zhu, En and Liu, Xinwang},
title = {Dynamic Weight-Enhanced Contrastive Learning for Partially View-Aligned Clustering},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2026},
volume = {36},
number = {8},
pages = {12426--12438},
doi = {10.1109/tcsvt.2026.3679096}
}
EUVIMED