TOM MITCHELL
Fredkin Professor of AI and Learning,
Chair, Machine Learning
www

I am interested in many areas of computer science, but especially in how to construct computers that learn from experience.  At the heart of the problem of machine learning is the question of how to automatically formulate general hypotheses given a collection of very specific training examples. My research has addressed a number of approaches to this question, including statistical approaches that find regularities over large numbers of training examples, and analytical approaches that generalize from very few examples and rely instead on prior knowledge and reasoning.

Much of my current research focuses around two projects:

Machine learning approaches to analyzing human brain activity. This project uses functional Magnetic Resonance Imaging (fMRI) to capture three-dimensional images of human brain activity at a spatial resolution of 1mm, once per second.  This is a wonderful set of data for studying the operation of the human brain, and because it is relatively new, there is a great need for new algorithms to analyze the data.  Recently we have demonstrated that it is possible to train machine learning algorithms to decode mental states of human subjects (e.g., to determine whether the word a person is examining is a noun or a verb) based on their observed fMRI brain activity.  I am interested in developing new algorithms that will help discover the spatial-temporal patterns of activity associated with a variety of brain processes, and that will help us better understand the working of the human brain.  We have access to the CMU-University of Pittsburgh  Brain Imaging Research Center, to design and collect data for our own experiments.

This project raises interesting machine learning questions such as how to train classifiers in extremely high dimensional, noisy data, and how to learn temporal models that characterize the evolution of hidden cognitive states while humans perform tasks such as reading and answering questions.

Intelligent workstation assistants that learn to help their users. This is part of a large multi-researcher project to build enduring, personalized, learning assistants for users of computer workstations (like us!).  We are working toward a software agent that can understand the user's email, calendar, text files, and actions, and that can learn the user's interests, habits, and tasks, in order to help in a wide range of activities.  My specific interest lies in how to make the agent learn.  For example, I am currently interested in the question of how the agent can learn to automatically extract information from text emails and files, and how it can learn what threads of activities the user is involved in, when, with whom, about what, etc.  This project raises many interesting machine learning questions about learning from labeled and unlabeled data, about learning and statistical language processing, and about cummulative learning over long periods of time.