AI Paper Atlas

Research map / Graph learning / GNN architectures

MLPInit: Embarrassingly Simple GNN Training Acceleration with MLP Initialization

Xiaotian Han, Tong Zhao, Yozen Liu, Xia Hu, Neil Shah

ICLR 2023 · Graph learning › GNN architectures · approach: Simple and scalable GNNs · 12 citations (OpenAlex)

Read the paper

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
Initialize GNNs with converged PeerMLP weights and fine-tuneGraphSAGE node classification · 7 node-classification datasetsaverage accuracy improvement: +7.97%Improved
Initialize GNNs with converged PeerMLP weights and fine-tuneGraphSAINT node classification · 7 node-classification datasetsaverage accuracy improvement: +7.00%Improved
Initialize GNNs with converged PeerMLP weights and fine-tuneClusterGCN node classification · 7 node-classification datasetsaverage accuracy improvement: +6.61%Improved
Initialize GNNs with converged PeerMLP weights and fine-tuneGCN node classification · 7 node-classification datasetsaverage accuracy improvement: +14.00% average; OGB-products: 77.08 -> 76.85Mixed
Initialize GraphSAGE with converged PeerMLP weightsGraphSAGE · OGB-productstraining speedup: 14.83×Improved
Initialize GCN with converged PeerMLP weightsGCN · OGB-productstraining speedup: 33.69×Improved
Initialize GraphSAGE with converged PeerMLP weightsGraphSAGE link prediction · PubMedHits@10: 28.38 -> 37.58Improved
Initialize GCN with converged PeerMLP weightsGCN link prediction · A-PhotoHits@10: 9.63 -> 9.17Hurt
Initialize GCN with converged PeerMLP weightsGCN link prediction · A-ComputersHits@10: 7.60 -> 6.76Hurt
Use PeerMLP weights without fine-tuning the GNNGCN · Cora node classificationaccuracy: 58.50 -> 77.60Improved
Use PeerMLP weights without fine-tuning the GNNGCN · CiteSeer node classificationaccuracy: 60.50 -> 69.70Improved
Use PeerMLP weights without fine-tuning the GNNGCN · PubMed node classificationaccuracy: 73.60 -> 78.10Improved

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