AI Paper Atlas

Research map / Graph learning / GNN architectures

InterpGNN: Understand and Improve Generalization Ability of Transdutive GNNs through the Lens of Interplay between Train and Test Nodes

Jiawei Sun, Kailai Li, Ruoxin Chen, Jie LI, Chentao Wu, Yue Ding, Junchi Yan

ICLR 2024 · Graph learning › GNN architectures · approach: Structural GNN variants

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Analyzes how information flow between train and test nodes affects generalization of transductive GNNs and proposes InterpGNN to improve it.

Problem

Transductive node prediction with GNNs often generalizes poorly due to limited information flow between distant nodes and intra-batch nodes in large graphs, and there were no formal theoretical results quantifying this.

Approach

Provides formal theoretical analysis of generalization in transductive GNNs via the interplay between train and test nodes, and uses these insights to design InterpGNN, a method to improve generalization.

Experimental setup

Builds on: message passing GNN

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy