Masked Multi-path Contrast with Confidence-Gated Semantic Imputation for Incomplete Multi-view Clustering

Incomplete multi-view clustering (IMVC) becomes particularly challenging under heavy missingness and view-availability imbalance. In this regime, scarce co-observed pairs make cross-view correspondences unreliable. Imputation-first pipelines may trigger cascading reconstruction errors, while purely consistency-based alignment often degrades sharply and gives limited control over semantic convergence across views. We propose MAGIC (Masked multi-pAth contrast with confIdence-Gated semantIc imputation), a unified framework that learns calibrated cluster semantics before conservative completion. MAGIC builds multiple correlated representation and prediction paths from lightly augmented latent codes, and couples them with a masked multi-path contrastive consensus objective and prediction-consistency regularization. The resulting posteriors are aggregated into view-wise soft assignments to reduce overconfidence and alleviate dominance by more frequently observed views. Based on these calibrated semantics, MAGIC performs similarity-guided semantic transfer in label space with confidence-aware gating, and completes missing representations through a geometry-preserving prototype fallback. Experiments on four benchmarks across different missing ratios show consistent gains over prior IMVC methods, and the ablations support the roles of masked multi-path consensus learning and confidence-gated semantic imputation.

This paper studies how to group data when each sample may have several different types of information, but some of them are missing. For example, a sample may have image, text, or sensor information, but not all views are available for every sample. This situation makes clustering difficult because the model may learn unreliable matches across views or spread early mistakes during completion. MAGIC addresses this problem by first learning more stable cluster information from multiple complementary paths, and then filling missing semantic information only when the model is confident enough. This cautious design reduces error propagation and improves clustering performance on several benchmark datasets.