Fine-to-Coarse Fairness-Informed Multi-View Clustering

In multi-view clustering (MVC), conventional anchor learning based models implicitly assume a uniform distribution of anchors across clusters, which could lead to inferior representation, especially when clusters vary significantly in size, as larger clusters require more anchors so as to adequately capture their intrinsic structural complexity. To alleviate this, we design a method termed FCFMVC that explicitly encourages proportional anchor allocation. To be specific, we transfer anchor allocation to discrete sample-cluster learning via bipartite graph bridge, and then backpropagate cluster state consisting of size and dispersion degree to guide anchor assignment. This allows the model to integrate cluster cardinality awareness and structural compactness directly into anchor distribution. On the other hand, we regard anchors as pseudo-samples, introduce an anchor-cluster indicator matrix on each view, and directly constrain the number of anchors assigned to each cluster within a tolerance margin. These two paths are further coupled through anchor-sample label alignment, and collaboratively facilitate anchor generation from fine-grained (anchor-level) to coarsegrained (cluster-level) structures. Besides, the entire optimization operation with linear time and space cost makes FCFMVC well-scalable to largescale tasks. Experiments on datasets with diverse scales confirm the effectiveness of our FCFMVC.

ORTF: This method extracts high-order inter-view correlations via non-negative tensor factorization, and utilizes orthogonal constraints to strengthen the discriminability of learned representations.

FMAC: This method introduces second order matching relationships to rearrange anchors, and builds anchor correspondences before information fusion to guarantee cross-view consistency.

LTBL: This method presents a low-rank tensor-driven proximity learning to model within-view and cross-view relationships, and employs tensor decomposition to retain high-dimensional structural information.

HOMGC: This method devises homophily principles to refine graph structures, and leverages dynamic graph adjustment to enhance intra-cluster connectivity and clustering performance. DGF: This method proposes a multi-graph learning to model consistent patterns and inconsistent noise, and exploits inter-view commonalities and view-specific unique properties to support comprehensive graph construction.

MFSA: This method combines feature selection and anchor graph factorization to extract informative representations, and utilizes word bag strategy to decrease feature redundancy.

UDBG: This method establishes a single optimization procedure to learn a unified anchor graph, and embeds graph generation and cluster learning to decrease the cumulative errors of separate two-stage processes. TTDM: This method captures intricate inter-view dependencies and intra-view structural patterns via tensorized tri-factor decomposition, and leverages the high-dimensional nature of tensor data to promote cluster structure discovery.