Linear Classifiers are Nearly Optimal When Hidden Variables Have Diverse Effect.

Nader H. Bshouty, Philip M. Long

Browse the full COLT paper archive.