IJCAI-2025-Consensus-Guided Incomplete Multi-view Clustering via Cross-view Affinities Learning (CAL)

Paper

This repository is the official matlab implementation for the paper "Consensus-Guided Incomplete Multi-view Clustering via Cross-view Affinities Learning" by Qian Liu (lq0307@dlmu.edu.cn), Huibing Wang (huibing.wang@dlmu.edu.cn), Jinjia Peng, Yawei Chen, Mingze Yao, Xianping Fu, Yang Wang. Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25.

Introduction

In this paper, we propose a novel Consensus-Guided Incomplete Multi-view Clustering via Cross-view Affinities Learning (CAL). Specifically, CAL reconstructs views with available instances to mine sample-wise affinities and harness comprehensive content information within views. Subsequently, to extract clean structural information, CAL imposes a structured sparse constraint on the representation tensor to eliminate biased errors. Furthermore, by integrating the consensus representation into a tensor stacked by view-specific affinity representations, CAL employs high-order interaction of multiple views to depict the semantic correlation between views while learning a unified structural graph across multiple views. Extensive experiments on several benchmark datasets demonstrate that CAL outperforms some state-of-the-art methods in clustering performance.

Dependencies

• OS: Windows 10
• Matlab2020b
• Related measure can be obtained in .\tools.

Data Preparation

• To generate the incomplete views, we randomly remove 10%, 30%, 50% and 70% samples of each view and at least preserve one view existing.
• For instance, we have provided the BBCSport dataset with a 10% missing data rate and the MSRC dataset with a 50% missing data rate in .\datasets, and the DataPreparing.m file is provided for data preparation.

Usage

• Conduct clustering run demoBBCSport.m and demoMSRC.m. The relevant hyperparameter settings have been provided in the code.

Citation

If any part of our paper and repository is helpful to your work, please generously cite with:

@inproceedings{ijcai2025p641,
  title     = {Consensus-Guided Incomplete Multi-view Clustering via Cross-view Affinities Learning},
  author    = {Liu, Qian and Wang, Huibing and Peng, Jinjia and Chen, Yawei and Yao, Mingze and Fu, Xianping and Wang, Yang},
  booktitle = {Proceedings of the Thirty-Fourth International Joint Conference on
               Artificial Intelligence, {IJCAI-25}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {James Kwok},
  pages     = {5761--5769},
  year      = {2025},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2025/641},
  url       = {https://doi.org/10.24963/ijcai.2025/641},
}