 |
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
|||||||||||
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
 |
 |
|
||||||||||
 |
|
|||||||||||
 |
|
|||||||||||
 |
 |
|
||||||||||
| Seminar Schedule (Seminar Organizer: Prof. Ziv Bar-Joseph) | |
Machine Learning Lunchtime Chats
|
Spring 2007 Seminar
Date: May 3, 2007 Time: 3:00 PM - 4:00 PM Location: 4615A Wean Hall Speaker: Steve Hanneke Ph.D. Candidate Title: The Complexity of Interactive Machine Learning Abstract: In supervised machine learning, well-designed interactive learning algorithms can provide significant reductions in the amount of effort required of human domain experts, compared to passive algorithms. But how much improvement can we expect for any given learning problem? In this talk, I will describe some recent progress toward formally answering this question. The talk will largely focus on the popular "label query" interactions, where the algorithm may request a class label for any example in a large pool of unlabeled data. I will describe general-purpose bounds on the number of label requests necessary and sufficient for learning, analogous to the sample complexity of passive learning. We will also consider generalizations to other types of interactions.