DGCNN: Disordered Graph Convolutional Neural Network Based on the Gaussian Mixture Model.

•We present a graph convolutional neural network (g-CNN) model that can perform feature learning on graph data directly.•We present the disordered graph convolutional layer that can change the discrete parameter learning into the sampling.•The experimental results show that the proposed method outperforms the existing g-CNN methods and other types of methods.