An Adaptive Framework for Multi-View Clustering Leveraging Conditional Entropy Optimization
Lijian Li
TL;DR
This paper tackles the robustness issue in multi-view clustering (MVC) arising from noisy views, by introducing CE-MVC, which jointly leverages conditional entropy and normalized mutual information to adaptively weight views and a parameter-decoupled deep model to learn view-specific representations. The core ideas are: (1) an asymptotic adaptive weighting scheme that quantifies view complementarity via conditional entropy and consistency via mutual information, (2) a per-view autoencoder-based architecture with independent parameters to prevent noise from leaking across views, and (3) iterative refinement of soft labels and a learning target to improve clustering. Empirical results demonstrate that CE-MVC consistently outperforms state-of-the-art MVC methods on synthetic noisy datasets and real-world data, with particular robustness gains in Noisy-DIGIT and Caltech. The approach offers a principled, scalable solution for reliable MVC in settings with heterogeneous and noisy perspectives, with potential impact on domains like medical imaging and multimodal data analysis.
Abstract
Multi-view clustering (MVC) has emerged as a powerful technique for extracting valuable insights from data characterized by multiple perspectives or modalities. Despite significant advancements, existing MVC methods struggle with effectively quantifying the consistency and complementarity among views, and are particularly susceptible to the adverse effects of noisy views, known as the Noisy-View Drawback (NVD). To address these challenges, we propose CE-MVC, a novel framework that integrates an adaptive weighting algorithm with a parameter-decoupled deep model. Leveraging the concept of conditional entropy and normalized mutual information, CE-MVC quantitatively assesses and weights the informative contribution of each view, facilitating the construction of robust unified representations. The parameter-decoupled design enables independent processing of each view, effectively mitigating the influence of noise and enhancing overall clustering performance. Extensive experiments demonstrate that CE-MVC outperforms existing approaches, offering a more resilient and accurate solution for multi-view clustering tasks.