AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Neural Relational Inference with Fast Modular Meta-learning

Ferran Alet, Erica Weng, Tomás Lozano-Pérez, Leslie Pack Kaelbling

NeurIPS 2019 · Graph learning › Knowledge and temporal graphs · approach: Heterogeneous graph learning

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
Replace variational independent edge prediction with joint simulated-annealing structure searchNRI · Charged particles edge type predictionaccuracy: 82.1 -> 88.4Improved
Replace variational encoder with modular meta-learningNRI · Springs edge type predictionaccuracy: 99.9 -> 99.9No clear effect
Replace variational NRI with modular meta-learningNRI · Charged particles, 10-step predictionMSE: 3.21e-3 -> 3.11e-3Improved
Replace variational NRI with modular meta-learningNRI · Springs, 10-step predictionMSE: 3.29e-6 -> 3.25e-6No clear effect
Use model-based inference instead of variational approach under small training setsNRI · Edge accuracy vs training set sizeaccuracyImproved
Use modular meta-learning edge inference instead of correlation of raw statesCorrelation(data) · Charged edge type predictionaccuracy: 55.8 -> 88.4Improved
Use LSTM hidden-state correlation instead of raw-state correlationCorrelation baseline · Springs edge type predictionaccuracy: 52.4 -> 52.7No clear effect
Use LSTM joint instead of single trajectoryLSTM · Springs 10-step predictionMSE: 4.69e-4 -> 2.19e-5Improved
Add a meta-learned proposal function to simulated annealing (ablation)Modular meta-learning · Search speedscalabilityImproved
Batch computation over many tasks using a super-graph (ablation)Modular meta-learning · Training and evaluation timespeedImproved
Add a latent unseen node optimized in the inner loopModular meta-learning with learned model · Springs with one particle hiddenMSE: 1.09e-3Improved
Use true graph instead of inferred graphNRI · Charged particles predictionMSE: 3.11e-3 -> 3.03e-3No 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