dlm@research.att.COM (11/25/87)
Date: November 20 (Friday) Time: 1:30 p.m. - 2:30 p.m. Place: AT&T Bell Labs Murray Hill 3D-473 Speaker: Leo Hartman Department of Computer Science University of Rochester Rochester, New York Performance in practical problem solving Abstract The quantity of resources that an agent expends in solving problems in a given domain is determined by the representations and search control strategies that it employs. The value of individual representations or strategies to the agent is determined by their contribution to the resource expenditure. We argue here that in order to choose the component representations and strategies appropriate for a particular problem domain it is necessary to measure their contribution to the resource expenditure on the actual problems the agent faces. This is as true for a system designer making such choices as it is for an autonomous mechanical agent. We present one way to measure this contribution and give an example in which the measure is used to improve problem solving performance. Sponsor: Henry Kautz