Learning Adaptive Neighborhoods for Graph Neural Networks

TL;DR

This paper introduces a differentiable graph generator that allows each node to learn its neighborhood and size, improving GCN performance especially when the original graph is noisy or missing.

Contribution

A novel end-to-end differentiable module for adaptive graph structure learning integrated into GCNs, enhancing flexibility and accuracy.

Findings

Improved accuracy in trajectory prediction, point cloud classification, and node classification.

Outperforms existing structure-learning methods across multiple datasets.

Applicable to any GCN architecture.

Abstract

Graph convolutional networks (GCNs) enable end-to-end learning on graph structured data. However, many works assume a given graph structure. When the input graph is noisy or unavailable, one approach is to construct or learn a latent graph structure. These methods typically fix the choice of node degree for the entire graph, which is suboptimal. Instead, we propose a novel end-to-end differentiable graph generator which builds graph topologies where each node selects both its neighborhood and its size. Our module can be readily integrated into existing pipelines involving graph convolution operations, replacing the predetermined or existing adjacency matrix with one that is learned, and optimized, as part of the general objective. As such it is applicable to any GCN. We integrate our module into trajectory prediction, point cloud classification and node classification pipelines resulting in improved accuracy over other structure-learning methods across a wide range of datasets and GCN backbones.