ATNPA: A Unified View of Oversmoothing Alleviation in Graph Neural Networks

arXiv:2405.01663v1 Announce Type: new
Abstract: Oversmoothing is a commonly observed challenge in graph neural network (GNN) learning, where, as layers increase, embedding features learned from GNNs quickly become similar/indistinguishable, making them incapable of differentiating network proximity. A GNN with shallow layer architectures can only learn short-term relation or localized structure information, limiting its power of learning long-term connection, evidenced by their inferior learning performance on heterophilous graphs. Tackling oversmoothing is crucial to harness deep-layer architectures for GNNs. To date, many …

abstract architectures arxiv become challenge cs.ai cs.lg embedding features gnn gnns graph graph neural network graph neural networks layer learn making network networks neural network neural networks them type view