Deep Graph Neural Networks with Shallow Subgraph Samplers
This work provides a significant improvement in accuracy and efficiency for researchers and practitioners working with deep GNNs on large graph datasets, overcoming common limitations like oversmoothing and neighborhood explosion.
This paper addresses the limitations of deep Graph Neural Networks (GNNs) regarding expressivity and computational cost. By proposing a "deep GNN, shallow sampler" design, the authors achieve state-of-the-art accuracy on ogbn-papers100M with an order of magnitude reduction in hardware cost.
While Graph Neural Networks (GNNs) are powerful models for learning representations on graphs, most state-of-the-art models do not have significant accuracy gain beyond two to three layers. Deep GNNs fundamentally need to address: 1). expressivity challenge due to oversmoothing, and 2). computation challenge due to neighborhood explosion. We propose a simple "deep GNN, shallow sampler" design principle to improve both the GNN accuracy and efficiency -- to generate representation of a target node, we use a deep GNN to pass messages only within a shallow, localized subgraph. A properly sampled subgraph may exclude irrelevant or even noisy nodes, and still preserve the critical neighbor features and graph structures. The deep GNN then smooths the informative local signals to enhance feature learning, rather than oversmoothing the global graph signals into just "white noise". We theoretically justify why the combination of deep GNNs with shallow samplers yields the best learning performance. We then propose various sampling algorithms and neural architecture extensions to achieve good empirical results. On the largest public graph dataset, ogbn-papers100M, we achieve state-of-the-art accuracy with an order of magnitude reduction in hardware cost.