From Neural Collapse to Label-Limited Evolving Streams: Geometry-Constrained Learning Under Dynamic Class Imbalance

Wang, Hongliang; Liu, Hongyuan; Hao, Qirui; Koh, Yun Sing; Yang, Qinli; Shao, Junming

doi:10.24963/ijcai.2026/341

From Neural Collapse to Label-Limited Evolving Streams: Geometry-Constrained Learning Under Dynamic Class Imbalance

Hongliang Wang, Hongyuan Liu, Qirui Hao, Yun Sing Koh, Qinli Yang, Junming Shao

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Main Track. Pages 3070-3078.

https://doi.org/10.24963/ijcai.2026/341

Keywords

Machine Learning: Active learning
Data Mining: Class imbalance and unequal cost
Data Mining: Mining data streams
Machine Learning: Semi-supervised learning

Abstract

Learning from label-limited streams presents significant challenges, particularly when coupled with concept drift and dynamic class imbalance. Existing works often struggle to maintain a discriminative feature space under these constraints, biasing decision boundaries toward majority classes or outdated concepts. To address this, we propose a novel framework named Neural collapse Inspired Label-limited Evolving stream learning (NILE). Instead of utilizing learnable classifiers, NILE exploits Neural Collapse (NC) geometry to explicitly construct a Simplex Equiangular Tight Frame (ETF) as a fixed classifier, ensuring maximal inter-class separability to guide feature discriminability. To maintain this discrimination with limited labels, NILE employs hybrid active learning to prioritize uncertain and minority samples, and an NC-adapted semi-supervised mechanism to enhance representation learning. NILE further updates the classifier by dynamically adjusting the ETF structure, enabling continual adaptation to drifting concepts. Extensive experiments show that NILE can effectively guide features toward the NC state, significantly outperforming state-of-the-art baselines in nonstationary environments.

Hongliang Wang, Hongyuan Liu, Qirui Hao, Yun Sing Koh, Qinli Yang, and Junming Shao. 2026. From Neural Collapse to Label-Limited Evolving Streams: Geometry-Constrained Learning Under Dynamic Class Imbalance. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 3070–3078. https://doi.org/10.24963/ijcai.2026/341

@inproceedings{ijcai2026p341,
  title     = {From Neural Collapse to Label-Limited Evolving Streams: Geometry-Constrained Learning Under Dynamic Class Imbalance},
  author    = {Wang, Hongliang and Liu, Hongyuan and Hao, Qirui and Koh, Yun Sing and Yang, Qinli and Shao, Junming},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {3070--3078},
  year      = {2026},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2026/341},
  url       = {https://doi.org/10.24963/ijcai.2026/341},
}

TY  - CPAPER
TI  - From Neural Collapse to Label-Limited Evolving Streams: Geometry-Constrained Learning Under Dynamic Class Imbalance
AU  - Wang, Hongliang
AU  - Liu, Hongyuan
AU  - Hao, Qirui
AU  - Koh, Yun Sing
AU  - Yang, Qinli
AU  - Shao, Junming
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 3070
EP  - 3078
DO  - 10.24963/ijcai.2026/341
UR  - https://doi.org/10.24963/ijcai.2026/341
ER  -