Confidence Graph Learning for Incomplete Multi-View Clustering

Abstract

Graph-based multi-view clustering has achieved significant attention due to its potent ability to represent clustering structures and robustness against noise. Due to human oversight or inherent factors in the application scenario, the collected multi-view data often has missing views. To achieve incomplete multi-view clustering, this paper proposes an effective confidence graph learning method, namely CGL-IMVC. Specifically, we propose a confidence graph learning scheme and further construct a confidence consensus graph. The confidence graph operates on an intuitive similar-nearest-neighbor hypothesis without additional information, thereby obtaining a high-quality consensus graph. A lot of experiments are conducted to validate the effectiveness of the proposed CGL-IMVC.