Factor Graph Neural Network
This work addresses the problem of modeling complex dependencies in structured data for machine learning applications, though it appears incremental as it builds on existing graph neural network frameworks.
The authors tackled the limitation of graph neural networks capturing only pairwise dependencies by generalizing them into a factor graph neural network (FGNN) to capture higher-order dependencies, showing it can represent Max-Product Belief Propagation and achieving promising results on synthetic and real datasets.
Most of the successful deep neural network architectures are structured, often consisting of elements like convolutional neural networks and gated recurrent neural networks. Recently, graph neural networks have been successfully applied to graph structured data such as point cloud and molecular data. These networks often only consider pairwise dependencies, as they operate on a graph structure. We generalize the graph neural network into a factor graph neural network (FGNN) in order to capture higher order dependencies. We show that FGNN is able to represent Max-Product Belief Propagation, an approximate inference algorithm on probabilistic graphical models; hence it is able to do well when Max-Product does well. Promising results on both synthetic and real datasets demonstrate the effectiveness of the proposed model.