108
click to vote
NIPS
2007
15 years 4 months ago
2007
We present a simple, agnostic active learning algorithm that works for any hypothesis class of bounded VC dimension, and any data distribution. Our algorithm extends a scheme of C...