Learning Discrete Structures for Graph Neural Networks
This addresses the limitation of GNNs for applications where graph data is imperfect or missing, enabling broader use in domains like social networks or bioinformatics, though it is an incremental improvement over existing GNN methods.
The paper tackles the problem that graph neural networks (GNNs) require graph structures, which are often noisy, incomplete, or unavailable in real-world scenarios, by proposing a method to jointly learn the graph structure and parameters of graph convolutional networks (GCNs), resulting in outperforming related methods by a significant margin.
Graph neural networks (GNNs) are a popular class of machine learning models whose major advantage is their ability to incorporate a sparse and discrete dependency structure between data points. Unfortunately, GNNs can only be used when such a graph-structure is available. In practice, however, real-world graphs are often noisy and incomplete or might not be available at all. With this work, we propose to jointly learn the graph structure and the parameters of graph convolutional networks (GCNs) by approximately solving a bilevel program that learns a discrete probability distribution on the edges of the graph. This allows one to apply GCNs not only in scenarios where the given graph is incomplete or corrupted but also in those where a graph is not available. We conduct a series of experiments that analyze the behavior of the proposed method and demonstrate that it outperforms related methods by a significant margin.