Entropy Aware Message Passing in Graph Neural Networks

Autor:	Nazari, Philipp, Lemke, Oliver, Guidobene, Davide, Gesp, Artiom
Rok vydání:	2024
Předmět:	Computer Science - Machine Learning I.2.6 I.5.1
Druh dokumentu:	Working Paper
Popis:	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. Comment: 4 pages, 3 figures
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2403.04636 Zobrazit plný text záznamu View this record from Arxiv