Incomplete Multi-View Clustering Based on Dual Aggregation Strategy and Dual Contrastive Completion

In real-world applications, data often comes from multiple views, and learning from multi-view data can improve clustering accuracy. While existing incomplete multi-view clustering methods have made progress in utilizing representation information, they often overlook global information, particularly raw data, which is seldom considered. To address this, we propose an incomplete multi-view clustering method based on dual aggregation and dual contrastive completion. Our approach aggregates both raw data and representations to capture global information, which is then combined with local view-specific information for contrastive learning and missing representation completion. The completed representations are used for clustering, with clustering results fed back to guide representation learning, enabling mutual optimization. Extensive experiments on three multi-view datasets demonstrate the effectiveness and robustness of our approach.