AI Paper Atlas

Research map / Graph learning / GNN architectures

Diverse Message Passing for Attribute with Heterophily

Liang Yang, Mengzhe Li, Liyang Liu, bingxin niu, Chuan Wang, Xiaochun Cao, Yuanfang Guo

NeurIPS 2021 · Graph learning › GNN architectures · approach: Graph rewiring and scalable message passing · 27 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
concatenation instead of summation for combining first/second-order information in DMP-1 (ablation)DMP-1 · Squirrel, Chameleon, Cornellaccuracy (%): 43.42 -> 47.26 (Squirrel)Improved
signed (tanh) propagation weights instead of positive softmax weights (ablation)DMP-1-Con · Squirrelaccuracy (%): 34.96 -> 47.26Improved
full per-attribute weights (DMP-1) instead of simplified DMP-2 (ablation)DMP Con · Wisconsinaccuracy (%): 84.31 -> 92.16Improved
per-attribute weights (DMP-1) instead of degenerated uniform DMP-Deg (ablation)DMP · Squirrelaccuracy (%): 32.46 -> 47.26Improved
DMP-1-Con instead of H2GCNheterophily networks · Squirrelaccuracy (%): 36.42 -> 47.26Improved
DMP-1-Con instead of GPRGNN or H2GCNhomophilous networks · Coraaccuracy (%): 88.14 (GPRGNN) vs 86.52No clear effect
stack more layers, DMP versus GCNDMP · Cora, Citeseer, Pubmed, Actoraccuracy vs depthImproved

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