DeeperGCN: All You Need to Train Deeper GCNs
This addresses a key bottleneck in graph representation learning for researchers and practitioners working with large-scale graphs, though it is an incremental improvement over existing GCN methods.
The paper tackled the problem of training deep Graph Convolutional Networks (GCNs), which suffer from issues like vanishing gradients and over-smoothing, by proposing DeeperGCN with novel components like differentiable aggregation functions and MsgNorm, resulting in significant performance boosts on large-scale graph learning tasks in the Open Graph Benchmark.
Graph Convolutional Networks (GCNs) have been drawing significant attention with the power of representation learning on graphs. Unlike Convolutional Neural Networks (CNNs), which are able to take advantage of stacking very deep layers, GCNs suffer from vanishing gradient, over-smoothing and over-fitting issues when going deeper. These challenges limit the representation power of GCNs on large-scale graphs. This paper proposes DeeperGCN that is capable of successfully and reliably training very deep GCNs. We define differentiable generalized aggregation functions to unify different message aggregation operations (e.g. mean, max). We also propose a novel normalization layer namely MsgNorm and a pre-activation version of residual connections for GCNs. Extensive experiments on Open Graph Benchmark (OGB) show DeeperGCN significantly boosts performance over the state-of-the-art on the large scale graph learning tasks of node property prediction and graph property prediction. Please visit https://www.deepgcns.org for more information.