 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
|||||||||||||
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
|||
 |
 |
 |
 |
|
||||||||||
 |
|
|||||||||||||
 |
|
|||||||||||||
 |
 |
|
||||||||||||
| Seminar Schedule (Seminar Organizer: Prof. Ziv Bar-Joseph) | |
Machine Learning Lunch Seminar
|
Fall 2009 Seminar
Date: October 15, 2009 Time: 4:30 PM - 00:00 AM () Location: CIC Bldg., Lower Level - Google Pittsburgh Other Speaker: Thorsten Joachims Associate Professor - Cornell University Title: CANCELED - Evaluating and Optimizing Search Engines through Interactive Experiments Abstract: The goal of a retrieval system is to provide the users with results of maximum utility. This raises two questions. First, how can utility be measured, and, second, how can it be optimized? The conventional approach is to equate utility with some score (e.g. Avg. Precision, NDCG) derived from expert relevance judgments, and then to optimize this score. However, this has several well-know problems (e.g. divergence between user and expert judgments, ignorance of user context, cost, availability), and it raises the question of whether utility can be elicited directly from the user? This talk will present methods for eliciting and optimizing utility directly via interactive experiments. Such interactive experiments give a more well-defined meaning to observable feedback like clicks, and they will be shown to provide accurate ordinal statements about result utility in a controlled user study. Furthermore, I will show how search engines can learn efficiently from a sequence of experiments in the sense of optimizing regret. This provides new methods for learning improved retrieval functions from implicit feedback. Joint work with Josef Broder, Bobby Kleinberg, Madhu Kurup, Filip Radlinski, and Yisong Yue. Speaker Bio: Thorsten Joachims is an Associate Professor in the Department of Computer Science at Cornell University. In 2001, he finished his dissertation with the title "The Maximum-Margin Approach to Learning Text Classifiers: Methods, Theory, and Algorithms", advised by Prof. Katharina Morik at the University of Dortmund. From there he also received his Diplom in Computer Science in 1997 with a thesis on WebWatcher, a browsing assistant for the Web. From 1994 to 1996 he was a visiting scientist at Carnegie Mellon University with Prof. Tom Mitchell. His research interests center on a synthesis of theory and system building in the field of machine learning, with a focus on Support Vector Machines, structured prediction, machine learning with text, and information retrieval. He authored the SVM-Light algorithm and software for support vector learning.
