The ability of a computer to learn by interacting with the environment is a fascinating subject. There are a few theoretical models of the learning process. The most important model is the Probably Approximately Correct (PAC) model: A learner gets examples labeled according to some target function. His purpose is to "learn" the target function, i.e., to output a hypothesis that approximates this function. In this seminar we will describe learning algorithms in this model, discuss learning in the presence of noise, and show how to boost weak learning algorithms to strong learning algorithms.
Lecture hours: | Monday 10:00-12:00. |
Reception hours: | Sunday 16:00-18:00, Building 58 (Math and CS building), Room 205. |
E-mail: | beimel at cs.bgu.ac.il |
Phone: | 647 7858 |
Course homepage: | www.cs.bgu.ac.il/~beimel/Courses/Learning2000/learning.html |