Computer Science Department




Resource-Bounded Reasoning Techniques

Spring 1997


Shlomo Zilberstein



Seminar Information

In many computer systems, it is not feasible (computationally) or desirable (economically) to compute the "optimal" answer. This seminar examines a variety of techniques that allow small quantities of computational commodities - such as time, memory, or information - to be traded for gains in the value of computed results. It covers both theory and applications in such areas as automated diagnosis and treatment, combinatorial optimization, probabilistic inference, and information gathering. Topics include: models of agent representation, decision theory and rational choice, the value of information, principles of meta-reasoning, memory-bounded search, utility-directed search, deliberation scheduling, dynamic planning and execution, and evaluation of resource-bounded reasoning techniques.

Time & place: Wednesday 10:30-1:00, LGRC 339

Prerequisites: Introduction to AI and probability theory or permission of instructor.

Credit: 3 units

Instructor: Shlomo Zilberstein,, 545-4189

Office hours: Monday 1:30-2:30, Thursday 11:00-12:00, LGRC A325

Each week we will discuss selected papers focused on a particular aspect of resource-bounded reasoning. Presentation of the material will be shared between instructor, seminar participants and occasional visitors. Seminar participants will be required to review papers, participate in discussions, make a few formal presentations, and complete a term project. The project can focus on an analytical study of a particular resource-bounded reasoning technique or on implementation and empirical evaluation. Participants will have to discuss their project with the instructor and turn-in a project proposal by 3/15/97. The final project report is due on 5/15/97.

Final grade will be based on class participation, material presentation, and project report.

A schedule of presentations, bibliography, and other handouts are available on-line at