An epoch-based improper active learning algorithm achieves realizability-level label complexity under a strictly weaker margin assumption on convex model classes.
Analysis of a greedy active learning strategy
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Active Learning via Regression Beyond Realizability
An epoch-based improper active learning algorithm achieves realizability-level label complexity under a strictly weaker margin assumption on convex model classes.