New bounds on the price of bandit feedback for mistake-bounded online multiclass learning.

Philip M. Long

Browse the full ALT paper archive.