Probabilistic and Regularized Graph Convolutional Networks

TL;DR

This paper reviews and reproduces key results of Graph Convolutional Networks, explores regularization and alternative approaches, and tests various configurations on standard datasets, finding limited improvements.

Contribution

It critically examines existing GCN methods, clarifies the role of regularization, and evaluates new configurations on benchmark datasets.

Findings

Explicit graph regularization was correctly rejected by original GCN.

Approximating a k-step transition matrix did not improve performance.

Various GCN configurations were tested on Cora, Citeseer, and Pubmed datasets.

Abstract

This paper explores the recently proposed Graph Convolutional Network architecture proposed in (Kipf & Welling, 2016) The key points of their work is summarized and their results are reproduced. Graph regularization and alternative graph convolution approaches are explored. I find that explicit graph regularization was correctly rejected by (Kipf & Welling, 2016). I attempt to improve the performance of GCN by approximating a k-step transition matrix in place of the normalized graph laplacian, but I fail to find positive results. Nonetheless, the performance of several configurations of this GCN variation is shown for the Cora, Citeseer, and Pubmed datasets.