Research in the Computational Learning Laboratory

Goals of the Computational Learning Laboratory

Computers have come to play a central role in many fields, one of the most important being the study of intelligent behavior. This includes work on learning and adaptation, in which experience leads to improvement on some task. In the last decade, advances in the computational study of this process have led to powerful insights about the nature of learning in both people and machines. The resulting computational techniques have led to an increasing number of applications within both industry and government.

However, one can apply the computational metaphor in different ways, and computational learning has become an important topic within many paradigms, including artificial intelligence, pattern recognition, control theory, cognitive psychology, and statistics. Such convergence of interests is encouraging, but few researchers in this active area communicate across disciplinary boundaries, and even fewer are skilled in the language and techniques of more than one approach. This leads to duplication of effort, repetition of conceptual errors, and missed insights that could come from interdisciplinary collaboration.

The Computational Learning Laboratory carries out a number of activities designed to bridge the gap that separates these paradigms. For example, it organizes a weekly seminar on computational learning and adaptation, which attracts scientists from a number of Stanford departments and local research centers. Talks address topics that hold general interest to learning researchers, presented in language accessible to people with different backgrounds.

Research in the Computational Learning Laboratory

In addition, the Laboratory's researchers are currently engaged in a number of interdisciplinary research efforts:

• Computational Discovery of Biological Models

• Computational Discovery of Quantitative Process Models

• Filtering Information in Complex Temporal Domains

• Modeling and Monitoring Organizational Behavior

• Data Mining in Spatio-temporal Domains.

• Adaptation and Learning in Autonomous Agents.

These activities involve collaborations with researchers at the Institute for the Study of Learning and Expertise, the Computational Science Division and the Center for Computational Astrobiology at NASA Ames Research Center, the Intelligent Communication Laboratory at the NTT Communication Science Laboratories in Kyoto, the Department of Plant Biology at the Carnegie Institution of Washington, the DaimlerChrysler Research and Technology Center, and various Stanford Departments.

In previous years, Laboratory researchers have contributed to other projects that address a variety of other challenging problems, including:

• Personalization for Route Guidance.

• Machine Learning for Image Analysis.

• Adaptive Interfaces for Destination Advice.

• Learning and Adaptation for Crisis Response.

• Machine Learning for Distributed Traffic Management.

Taken together, these projects deal with many facets of intelligent behavior, and combine ideas from a variety of disciplines, including artificial intelligence, decision theory, and cognitive psychology. However, they share a concern with the computational nature of learning and adaptation, and they incorporate a common bias toward testing theoretical ideas in realistic settings. In the future, the Computational Learning Laboratory plans to pursue additional interdisciplinary efforts with similar characteristics.

Recent Papers