Scholarlink

Gated Graph Neural Network (GGNN) is a type of Graph Neural Network (GNN) that introduces the gating mechanism to control the flow of information between nodes. In classic GNN, each node receives messages from its neighbors and updates its representation based on them. However, not all messages are equally important. The gating mechanism in GGNN modulates how much information from its neighbors each node should integrate into its feature. In this work, we show the impact of the introduction of gating mechanism in the GNN modifying the generic node of the graph, and focusing on classification task. This also allows us to obtain a visualization method that can improve our comprehension of gating in GNNs.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gated Graph Neural Networks for Classification Task

  • 章节

  • 发表日期：

    2025年

查看全文
  • Amedeo Buonanno,
  • Giovanni Di Gennaro,
  • Armando Ospedale,
  • Francesco A. N. Palmieri

摘要

Gated Graph Neural Network (GGNN) is a type of Graph Neural Network (GNN) that introduces the gating mechanism to control the flow of information between nodes. In classic GNN, each node receives messages from its neighbors and updates its representation based on them. However, not all messages are equally important. The gating mechanism in GGNN modulates how much information from its neighbors each node should integrate into its feature. In this work, we show the impact of the introduction of gating mechanism in the GNN modifying the generic node of the graph, and focusing on classification task. This also allows us to obtain a visualization method that can improve our comprehension of gating in GNNs.