4 Federated Personalized Subgraph Learning 4.1 Subgraph Similarity Estimation Model Parameters for Subgraph Similarities(参数相似度) 首先考虑使用模型参数之间的相似度进行度量,即 S=\frac{\theta_i \cdot \theta_j}{\Vert \theta_i\Vert\Vert \theta_j\Vert} ,即余弦相似度。但由于余弦相似度受到维数灾难的...
论文链接:Personalized Subgraph Federated Learning (mlr.press) 代码链接:GitHub - JinheonBaek/FED-PUB: Official Code Repository for the paper - Personalized Subgraph Federated Learning (ICML 2023) 1、引言 大多数现有的图神经网络(GNNs)都着重于单个图,其节点和边从多个来源收集,并存储在中央服务器中。然而...