Graph Wavelet Convolutional Network with Graph Clustering

In this paper, we present a novel Graph Wavelet Convolutional Network (GWCN) approach with a graph clus-tering algorithm such as METIS. GWCN is a graph wavelet transform-based method. It has better locality than Graph Convolutional Network (GCN) using the graph Fourier transform, and results higher classification accuracy. In this work, the graph clustering algorithm is applied to GWCN for providing a mechanism to the mini-batch selection in deep learning, which has an effective impact on learning.