Convolutional Neural Network Architectures for Signals Supported on Graphs

Abstract

Two architectures that generalize convolutional neural networks (CNNs) for the processing of signals supported on graphs are introduced. We start with the selection graph neural network (GNN), which replaces linear time invariant filters with linear shift invariant graph filters to generate convolutional features and reinterprets pooling as a possibly nonlinear subsampling stage where nearby nodes pool their information in a set of preselected sample nodes. A key component of the architecture is to remember the position of sampled nodes to permit computation of convolutional features at deeper layers. The second architecture, dubbed aggregation GNN, diffuses the signal through the graph and stores the sequence of diffused components observed by a designated node. This procedure effectively aggregates all components into a stream of information having temporal structure to which the convolution and pooling stages of regular CNNs can be applied. A multinode version of aggregation GNNs is further introduced for operation in large scale graphs. An important property of selection and aggregation GNNs is that they reduce to conventional CNNs when particularized to time signals reinterpreted as graph signals in a circulant graph. Comparative numerical analyses are performed in a source localization application over synthetic and realworld networks. Performance is also evaluated for an authorship attribution problem and text category classification. Multinode aggregation GNNs are consistently the best performing GNN architecture.

HodgeNet: Graph Neural Networks for Edge Data

“…Following the experiments in [13], we consider a graph drawn from a planted partition model with p = 0.8, q = 0.2, and k = 5 communities of 20 nodes each. A set of 10000 training signals is then generated by randomly choosing a diffusion time 1 ≤ t i ≤ 20 uniformly at random for each signal indexed by i ∈ {1, 2, .…”

“…. , 10000}, then applying the diffusion operator D ti (A) = (A/λ 1 ) ti to an impulse δ vi as in (13), where v i is randomly chosen among the nodes of highest degree in each community. Here, λ 1 denotes the largest eigenvalue of A.…”

Optimal Power Flow Using Graph Neural Networks

“…In this work, we use imitation learning and graph neural networks [13][14][15] to find a local and scalable solution to the OPF problem. More specifically, we adopt a parametrized GNN model, which is local and scalable by design, and we train it to imitate the optimal solution obtained using an interior point solver [16], which is centralized and does not converge for large networks.…”

Towards an Efficient and General Framework of Robust Training for Graph Neural Networks

“…Graph structured representations are one of the most commonly encountered data structure that naturally arises in nearly all scientific and engineering application [1]. With the widespread use of networks neuroscience, molecular chemistry and other fields, it is not surprising that machine learning on graph data has become a key learning tool.…”