AI Paper Atlas

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

Debugging and Explaining Metric Learning Approaches: An Influence Function Based Perspective

Ruofan Liu, Yun Lin, XIANGLIN YANG, Jin Song Dong

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 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
replace the original influence function with the empirical influence function (EIF)influence-based debugging of DML · CUB200 individual confusion pairsaverage distance improvement: 2.29E-02 at top-30 pairs, p-value 3.47E-12Improved
replace IF with EIFinfluence-based debugging of DML · CARS196 individual confusion pairsaverage distance improvement: 1.47E-02 at top-30 pairs, p-value 8.16E-06Improved
replace IF with EIFinfluence-based debugging of DML · InShop individual confusion pairsaverage distance improvement: 1.52E-02 at top-30 pairs, p-value 1.96E-05Improved
replace IF with EIF on groups of confusion pairsinfluence-based debugging of DML · InShop group confusionsignificance of improvementMixed
use EIF instead of IF for mislabelled sample detectioninfluence-based debugging of DML · CUB200runtime (s): 375.56 ± 68.06 -> 260.20 ± 29.28Improved
use EIF instead of IFinfluence-based debugging of DML · CARS196runtime (s): 588.11 ± 23.23 -> 453.61 ± 28.76Improved
use EIF instead of IFinfluence-based debugging of DML · InShopruntime (s): 2454.11 ± 9.55 -> 1306.43 ± 27.31Improved
raise mislabelling ratio from 1% to 10%EIF relabelling on ProxyNCA++ · CUB200relabelling recommendation performance: 75.68% -> 83.28%Improved
raise mislabelling ratio from 1% to 10%EIF relabelling on SoftTriple · InShoprelabelling recommendation performance: 90.00% -> 86.27%Hurt
raise mislabelling ratio from 1% to 10%EIF relabelling on SoftTriple · CARS196relabelling recommendation performance: 79.49% -> 89.95%Mixed
vary number of sampled parameters N_theta in EIFEIF · CUB200, CARS196, InShopratio of mislabelled samples detectedNo clear effect
rank samples by influence magnitude, keeping only small-magnitude scoresEIF noisy-sample detection · 3 DML datasetsratio of mislabelled samples detectedHurt

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