AI Paper Atlas

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

Adaptive Neighborhood Metric Learning

Kun Song, Junwei Han, Gong Cheng, Jiwen Lu, Feiping Nie

TPAMI 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 34 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
add neighborhood parameters (lambda, gamma) to loss (ablation)Lifted structure loss · CUB-200-2011Recall@1: 45.4 -> 47.1Improved
add neighborhood parameters (lambda, gamma) to loss (ablation)Proxy-NCA · CUB-200-2011Recall@1: 49.2 -> 52.7Improved
add neighborhood parameters (lambda, gamma) to loss (ablation)N-pairs loss · CUB-200-2011Recall@1: 43.6 -> 46.8Improved
add neighborhood parameters (lambda1, lambda2) (ablation)DANML · CUB-200-2011Recall@1: 65.4 -> 67.6Improved
replace multi-similarity loss with DANMLMulti-similarity loss · CUB-200-2011Recall@1: 65.7 -> 67.6Improved
replace multi-similarity loss with DANMLMulti-similarity loss · Cars-196Recall@1: 84.1 -> 85.6Improved
replace Circle Loss with DANMLCircle Loss · Cars-196Recall@1: 83.4 -> 85.6Improved
use adaptive neighborhood instead of fixed 10-nearest-neighbor targets (LANML-)LMNN · 15 UCI datasetsRecall@1Improved
add parameter alpha to NCA (PNCA)NCA · 15 UCI datasetsRecall@1Improved
use ANML with a continuous log-exp mean surrogateLMNN · running timeRecall@1Improved

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