Research Interests

My research is focused using machine learning to learn from data collected implicitly from web users. My interests include personalized search, learning to rank, advertising as an optimization problem and how to leverage the internet to help students learn.

My publications are available here.

Learning from Implicit Feedback

Web usage logs provide a wealth of information about user needs, document collections and search performance. However this data is highly biased and noisy. I have explored techniques to interpret usage logs, effective machine learning approaches that use this data, and addressed the question of how to treat learning from usage logs as an active learning problem.

Learn to Rank

Rankings are a very common format for presenting items to human beings. I am interested in formulations of ranking as an optimization problem, finding appropriate loss metrics and making most effective use of available data for learning to rank. Moreover, in the context of the web search, effective ranking is needed both for web search results and online advertisements. This leads to two quite different optimization problems that I have been exploring in my research.

Computer Assisted Education

The Internet provides a unique opportunity to study how students use self-teaching resources by observing their interactions with a web browser for very large numbers of students. I am interesting in using implicitly collected information about such user behavior to improve resources available to students.

Older Research

Evolutionary Methods for Learning on Structured Data (Undergraduate work).

Cornell Research Links