Entropy Aware Message Passing in Graph Neural Networks

arXiv:2403.04636v1 Announce Type: new
Abstract: Deep Graph Neural Networks struggle with oversmoothing. This paper introduces a novel, physics-inspired GNN model designed to mitigate this issue. Our approach integrates with existing GNN architectures, introducing an entropy-aware message passing term. This term performs gradient ascent on the entropy during node aggregation, thereby preserving a certain degree of entropy in the embeddings. We conduct a comparative analysis of our model against state-of-the-art GNNs across various common datasets.

abstract aggregation architectures arxiv cs.lg entropy gnn gradient graph graph neural networks issue networks neural networks node novel paper physics struggle type