Dynamic fusion and non-negative matrix factorization-based multiview clustering method

Aiming at the problem that most of the current multi-view clustering methods only focus on the information of a single view, ignoring the correlation between views, and failing to fully explore the potential structure of the data, a multi-view clustering method combining dynamic fusion and non-negative matrix factorization is proposed. First, we extract low-dimensional feature representations for each view using non-negative matrix factorization and enhance feature consistency across different views through contrastive learning. Second, we apply a continual learning strategy to dynamically update the model, adapting to changes in new view data and improving model robustness and adaptability. Finally, we use a dynamic fusion strategy to weight and integrate features from all views, obtaining a common feature representation for multi-view data. Experimental results demonstrate that our proposed method significantly outperforms existing methods in clustering performance on several public datasets.