Yoav S. Freund
Professor, Computer Science and Engineering
Faculty, Office of the Dean of Engineering
Computational learning theory and related areas in probability theory, information theory, statistics and pattern recognition. Yoav Freund works on applications of machine learning algorithms in bioinformatics, computer vision, finance, network routing and high-performance computing. He has developed a new approach to the study and development of machine learning algorithms, where the goal is to produce a good decision algorithm for a repetitive decision task. A decision algorithm receives as input an instance (sensory data) and outputs a decision (an action). After the decision has been made, there is a measurable outcome. Freund's main focus is on binary classification tasks, where the decision is binary, the outcome is binary, and the loss is 1 if the decision and outcome don't match and 0 if they do. Given these definitions and a source of instances and outcomes, Freund can evaluate the performance of any decision algorithm--treating it as a "black box". The practical advantage of the black-box approach is that it provides a measuring stick for comparing all types of decision algorithms, regardless of how they are constructed or analyzed. By extension, this approach produces a way of comparing all types of learning algorithms.
Jacobs School Faculty Update Your Profile