Incomplete multi-view classification via graph neural network on heterogeneous graph

Guoqing Chao, Jingnan Qi, Jingxuan Li, Bumshik Lee, Dianhui Chu

Abstract: Highlights•We proposed a novel incomplete multi-view learning method with heterogeneous graph.•Two mutual information losses are designed to exploit consensus and complementarity.•Experiments verified the effectiveness and superiority of the proposed method.