Diameter-Based Active Learning

Abstract

To date, the tightest upper and lower-bounds for the active learning of general concept classes have been in terms of a parameter of the learning problem called the splitting index. We provide, for the first time, an efficient algorithm that is able to realize this upper bound, and we empirically demonstrate its good performance.

16 pages, 2 figures

References in corpus (1)

• Average-Case Active Learning with Costs

Cited by in corpus (3)

• Flattening a Hierarchical Clustering through Active Learning
• Distribution Matching for Machine Teaching
• A Structured Perspective of Volumes on Active Learning