Systems Seminar

Agnostic Active Learning

Dr. John Langford
Toyota Technological Institute at the University of Chicago

Abstract

On the positive hand: It is well known that active learning can result in exponential reductions in the number of labeled samples required to successfully learn a concept.

On the negative hand: An examination of typical active learning algorithms results in the discovery that they are not robust against noise. It is easy to imagine that noise robustness is critical to real- world performance.

A fundamental question is: Can we achieve the exponential improvements of active learning in a robust fashion? The answer is "yes", and I'll explain how.

Time and Place: Tue., Nov. 8, at 1:30 pm in 3609 Engr. Hall.       *** NOTE SPECIAL DAY, TIME, & ROOM ***

SYSTEMS SEMINAR WEB PAGE: http://homepages.cae.wisc.edu/~gubner/seminar/

File "langford.shtml" last modified Tue 15 Oct 2019, 01:45 PM, CDT
Web Page Contact: John (dot) Gubner (at) wisc (dot) edu