Fisher-Bures Adversary Graph Convolutional Networks

In a graph convolutional network, we assume that the graph G is generated wrt some observation noise. During learning, we make small random perturbations DeltaG of the graph and try to improve generalization. Based on quantum information geometry, DeltaG can be characterized by the eigendecomposition of the graph Laplacian matrix. We try to minimize the loss wrt the perturbed G+DeltaG while making DeltaG to be effective in terms of the Fisher information of the neural network. Our proposed model can consistently improve graph convolutional networks on semi-supervised node classification tasks with reasonable computational overhead. We present three different geometries on the manifold of graphs: the intrinsic geometry measures the information theoretic dynamics of a graph; the extrinsic geometry characterizes how such dynamics can affect externally a graph neural network; the embedding geometry is for measuring node embeddings. These new analytical tools are useful in developing a good understanding of graph neural networks and fostering new techniques.