AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Learning Large Graph Property Prediction via Graph Segment Training

Kaidi Cao, Phitchaya Mangpo Phothilimthana, Sami Abu-El-Haija, Dustin Zelle, Yanqi Zhou, Charith Mendis, Jure Leskovec, Bryan Perozzi

NeurIPS 2023 · Graph learning › Knowledge and temporal graphs · approach: Graph reasoning for visual recognition · 1 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
train with graph segments instead of full graphGCN · MalNet-Tinytest accuracy: 87.84 -> 88.26No clear effect
train on only one random segment per graph (GST-One) (ablation)GST with SAGE · MalNet-Largetest accuracy: 88.62 -> 57.13Hurt
add historical embedding table alone (GST+E) (ablation)GST with GCN · MalNet-Largetest accuracy: 88.35 -> 48.42Hurt
add prediction head finetuning to GST+E (ablation)GST+E with SAGE · MalNet-Largetest accuracy: 43.28 -> 85.26Improved
add stale embedding dropout (SED) to GST+E (ablation)GST+E with SAGE · MalNet-Largetest accuracy: 43.28 -> 71.83Improved
add SED to GST+EF (full GST+EFD) (ablation)GST+EF with SAGE · MalNet-Largetest accuracy: 85.26 -> 89.78Improved
GST+EFD instead of GSTGCN · MalNet-Largetest accuracy: 88.35 -> 89.67Improved
GST+EFD instead of GSTGraphGPS · MalNet-Tinytest accuracy: 91.03 -> 92.46Improved
GST+EFD instead of GSTTpuGraphs · TpuGraphstest OPA: 86.84 -> 89.20Improved
GST+EFD instead of GST: lower train OPATpuGraphs · TpuGraphstrain OPA: 87.43 -> 71.40Hurt
GST+EFD instead of GST for runtimeGCN · MalNet-Largetraining time per iteration (ms): 720.8 -> 247.9Improved
SED embedding fetch overhead: GST+EFD vs GST-One runtimeGCN · MalNet-Largetraining time per iteration (ms): 242.2 -> 247.9No clear effect

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