AI Paper Atlas

Research map / Graph learning / GNN architectures

LD2: Scalable Heterophilous Graph Neural Network with Decoupled Embeddings

Ningyi Liao, Siqiang Luo, Xiang Li, Jieming Shi

NeurIPS 2023 · Graph learning › GNN architectures · approach: Simple and scalable GNNs

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
use LD2 decoupled precomputation instead of LINKXLD2 · geniustest accuracy (%): 82.51 -> 85.31Improved
use LD2 instead of LINKXLD2 · pokectest accuracy (%): 68.64 -> 74.93Improved
use LD2 instead of LINKXLD2 · snap-patentstest accuracy (%): 52.69 -> 58.58Improved
use LD2 instead of LINKXLD2 · penn94test accuracy (%): 78.63 -> 75.52Hurt
use LD2 instead of LINKXLD2 · arxiv-yeartest accuracy (%): 50.44 -> 50.29No clear effect
add graph topology embeddings on top of node features (LD2 vs MLP)LD2 · snap-patentstest accuracy (%): 23.03 -> 58.58Improved
use LD2 instead of LINKX for GPU memoryLD2 · snap-patentspeak GPU memory (GB): 21.53 -> 3.96Improved
use LD2 instead of LINKX for GPU memoryLD2 · wikipeak GPU memory (GB): 14.53 -> 4.47Improved
decoupled precomputation vs heterophilous baselines in training timeLD2 · large graphstraining time: 3-15 times fasterImproved
apply GSAINT minibatch sampling to MixHopMixHop-GS · snap-patentstest accuracy (%): 34.73 vs LD2 58.58Hurt
add feature embeddings to raw attributes on genius (ablation)LD2 channels · geniustest accuracy (%)Improved
use only adjacency embedding PA vs integrating with other channels (ablation)LD2 channels · geniustest accuracy (%)Improved

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