AI Paper Atlas

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

A Deep Model With Local Surrogate Loss for General Cost-Sensitive Multi-Label Learning

Cheng-Yu Hsieh, Yi-An Lin, Hsuan-Tien Lin

AAAI 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning · 12 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
fine-tune with LLSL instead of BCE lossBCE-trained deep model · Bibtexnormalised P@1: 0.4482 -> 0.3647Hurt
fine-tune with LLSL instead of BCE lossBCE-trained deep model · Deliciousnormalised P@3: 0.3778 -> 0.3581Hurt
fine-tune with LLSL instead of BCE lossBCE-trained deep model · Wiki10-31Knormalised P@5: 0.3631 -> 0.3586Hurt
use LLSL instead of SLEECdeep model · Bibtexnormalised P@3: 0.6036 -> 0.6157Improved
use LLSL instead of SLEECdeep model · Deliciousnormalised P@3: 0.3862 -> 0.3581Hurt
use weighted BCE instead of BCEdeep model · Example-F1 and Rank lossrelative rankingImproved
use LLSL instead of weighted BCEdeep model · example-based criteriarelative rankingImproved
use polynomial two-bit local approximator instead of linear one-bit (ablation)LLSL · Example-F1, Micro-F1, Macro-F1relative rankingImproved
use linear one-bit local approximator instead of more complex ones (ablation)LLSL · Hamming loss and Rank lossrelative rankingImproved
mix BCE (Hamming) loss with learned surrogate for Precision-at-k (ablation)LLSL · many-label datasetsconvergenceImproved

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