Fig. 2

An illustration of convolution operation and graph convolution operation. (a) The results of the standard convolution operation are the sum of the point-wise multiplications of a subset of the signal \(\:\mathbf{X}\) with a learned convolution filter \(\:\mathbf{W}\). (b) The results of graph convolution operation could be the sum of the node feature itself with the product of the features of adjacent nodes \(\:{\mathbf{h}}_{\mathbf{N}\mathbf{e}}\) and the learnable parameter \(\:\mathbf{W}\).