Modification of Architecture Learning Convolutional Neural Network for Graph

PATCHY-SAN is a framework for learning Convolutional Neural Network (CNNs) for an arbitrary graph. In this paper, we propose a modified architecture of Convolutional Neural Network in PATCHY-SAN. We use some representation matrices of a graph such as B-i, L-i, N-i, with B, L, N, are a betweenness matrix, a Laplacian matrix and a normalize Laplacian matrix with i = 1, 2, 3, 4, 5. We do some experiment of a model with three convolutional layers and two convolutional layers. In order to reduce internal covariate shift, we use a batch normalization as a regularizer. In conclusion, by adding more convolution layers, and using batch normalization can increase and reduce accuracy. The accuracy is more dependent on the architecture of CNNs.