Incomplete Multi-view Clustering via Prototype-based Imputation

  • 2023-01-30 03:52:24
  • Haobin Li, Yunfan Li, Mouxing Yang, Peng Hu, Dezhong Peng, Xi Peng
  • 0


In this paper, we study how to achieve two characteristics highly-expected byincomplete multi-view clustering (IMvC). Namely, i) instance commonality refersto that within-cluster instances should share a common pattern, and ii) viewversatility refers to that cross-view samples should own view-specificpatterns. To this end, we design a novel dual-stream model which employs a dualattention layer and a dual contrastive learning loss to learn view-specificprototypes and model the sample-prototype relationship. When the view ismissed, our model performs data recovery using the prototypes in the missingview and the sample-prototype relationship inherited from the observed view.Thanks to our dual-stream model, both cluster- and view-specific informationcould be captured, and thus the instance commonality and view versatility couldbe preserved to facilitate IMvC. Extensive experiments demonstrate thesuperiority of our method on six challenging benchmarks compared with 11approaches. The code will be released.


Quick Read (beta)

loading the full paper ...