AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Discriminative Gaifman Models

Mathias Niepert

NeurIPS 2016 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 7 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
Increase neighborhood depth r from 1 to 3 (with k=20, w=5, w̃=25) (ablation)Gaifman model · FB15KHits@10: 82.1 -> 84.2Improved
Increase neighborhood depth r from 1 to 2 (k=20, w=5, w̃=25) (ablation)Gaifman model · WN18Hits@10: 93.6 -> 93.9Improved
Use more neighborhood samples at inference (N=1 to 3 averaged) and more training neighborhoods (w=1,w̃=2 to w=5,w̃=25) (ablation)Gaifman model · FB15KHits@1: 60.1 -> 65.6Improved
Use more training neighborhood samples (w=1,w̃=2 to w=5,w̃=25) with k=20, r=1 (ablation)Gaifman model · WN18Hits@10: 88.1 -> 93.6Improved
Remove size bound k (use full 1-neighborhood) instead of k=20 (ablation)Gaifman model · FB15KHits@10: 82.1 -> 78.1Hurt
Remove size bound k (use full 1-neighborhood) instead of k=20 (ablation)Gaifman model · WN18Hits@10: 93.9 vs 93.6No clear effect
Gaifman model with k=20 bounded neighborhoods vs TransE baselineGaifman model · FB15KHits@1: 28.1 -> 65.6Improved
Use neighborhood-based Gaifman models (r=1, k=20) instead of DistMult on WN18Gaifman model · WN18Hits@10: 93.6 vs 93.7Mixed

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