Tutorial 3: Ranking Methods in Machine Learning


Ranking problems are increasingly recognized as a new class of statistical learning problems that are distinct from the classical learning problems of classification and regression. Such problems arise in a wide variety of domains: in information retrieval, one wants to rank documents according to relevance to a query; in natural language processing, one wants to rank alternative parses or translations of a sentence; in collaborative filtering, one wants to rank items according to a user’s likes and dislikes; in computational biology, one wants to rank genes according to relevance to a disease. Consequently, there has been much interest in ranking in recent years, with a variety of methods being developed and a whole host of new applications being discovered. This tutorial aims to make these recent advances in ranking methods in machine learning accessible to a wider audience. In particular, participants will learn to distinguish between different forms of ranking problems (such as instance ranking, label ranking, subset ranking, rank aggregration, etc), to identify the form of ranking problem most suited for a given application, and to determine what learning methods are most appropriate for a given form of ranking problem.

Tutor's Biography:

Shivani Agarwal is a postdoctoral associate in the Computer Science and Artificial Intelligence Laboratory at MIT. She obtained her PhD in Computer Science at the University of Illinois, Urbana-Champaign, where she received the Liu Award for her research; a BA with Honors in Computer Science at Trinity College, University of Cambridge, where she was a Nehru Scholar; and a BSc with Honors in Mathematics at St Stephen's College, University of Delhi. Her research interests include machine learning and learning theory, in particular the study of ranking and other new learning problems, as well as applications of machine learning methods, particularly in the life sciences. More broadly, she is excited by research at the intersection of computer science, mathematics, and statistics, and its applications in scientific discovery.

The tutor has published extensively in the area of ranking in machine learning, and has organized two workshops related to the topic: an early workshop titled ‘Learning to Rank’ at NIPS 2005, and a workshop titled ‘Advances in Ranking’ that is being organized this year at NIPS 2009.