AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Temporal Graph Benchmark for Machine Learning on Temporal Graphs

Shenyang Huang, Farimah Poursafaei, Jacob Danovitch, Matthias Fey, Weihua Hu, Emanuele Rossi, Jure Leskovec, Michael M. Bronstein, Guillaume Rabusseau, Reihaneh Rabbany

NeurIPS 2023 · Graph learning › Knowledge and temporal graphs · approach: Continuous-time dynamic graph learning · 22 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
Use persistence forecast instead of a temporal graph modelNode affinity prediction · tgbn-tradeNDCG@10: 0.374 -> 0.855 (DyRep); 0.374 -> 0.855 (TGN)Improved
Use moving average instead of a temporal graph modelNode affinity prediction · tgbn-genreNDCG@10: 0.351 -> 0.509 (DyRep); 0.367 -> 0.509 (TGN)Improved
Use moving average instead of a temporal graph modelNode affinity prediction · tgbn-redditNDCG@10: 0.312 -> 0.559 (DyRep); 0.315 -> 0.559 (TGN)Improved
Use moving average instead of a temporal graph modelNode affinity prediction · tgbn-tokenNDCG@10: 0.141 -> 0.508 (DyRep); 0.169 -> 0.508 (TGN)Improved
Increase negative samples from 20 to all possible destinations (ablation)Dynamic link prediction on tgbl-wiki · tgbl-wikiMRR: 0.791 -> 0.711Hurt
Increase negative samples from 20 to 100 per positive (ablation)Dynamic link prediction on tgbl-review · tgbl-reviewMRR: 0.532 -> 0.349Hurt
Apply the TGB evaluation across diverse datasets rather than relying on tgbl-wiki aloneTemporal link prediction models · tgbl-wiki and tgbl-reviewMRR: NAT: 0.749 -> 0.341; GraphMixer: 0.118 -> 0.521Mixed
Use a simple memory heuristic rather than DyRepDynamic link prediction · tgbl-coinMRR: 0.452 -> 0.580Improved

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