Multi-stage cost-efficient multi-label active learning

doi:10.1016/j.asoc.2026.115445

Multi-stage cost-efficient multi-label active learning

2026-05-14

0

PRE

AI

DOI：10.1016/j.asoc.2026.115445

Original

Original request for help

Save

Abstract

En 中文

• A novel multi-stage cost-efficient multi-label active learning strategy known as MCMAL is proposed. • A cooperative mechanism is developed that combines pseudo-labeling and manual annotation to label low-value sample pairs. • Experiments demonstrate that MCMAL makes cost-effective decisions and reduces annotation costs for multi-label samples. • In the multi-stage active learning strategy, the annotation value measurement methods at each stage are interchangeable.

Keywords:

MCMAL

multi-label active learning

cost-efficient

pseudo-labeling

annotation value measurement

Multi-stage cost-efficient multi-label active learning

Abstract

Journal

Organization

Cited Papers