Through the introduction of a diffusion-convolution operation, we show how diffusion-based representations can be learned from graph-structured data and used as an effective basis for node classification. DCNNs have several attractive qualities, including a latent representation for graphical data that is invariant under isomorphism, as well as polynomial-time prediction and learning that can be represented as tensor operations and efficiently implemented on a GPU. Through several experiments with real structured datasets, we demonstrate that DCNNs are able to outperform probabilistic relational models and kernel-on-graph methods at relational node classification tasks.

The idea of incorporating graph diffusion into neural networks seem both interesting and novel. The authors also did a good job in motivating the problem. However, overall I feel several aspects of the work could be further improved: Scalability: 1. The authors proposed three separate models for node, graph and edge classification. However, no empirical performance of edge classification was reported.