Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Tuesday, January 16, 2007, 12:15 pm
Duration: This information is not available in the database
Location: CAB G51
Speaker: Leslie Valiant (Division of Engineering and Applied Sciences, Harvard Univ.)
It has been long recognized that cognitive phenomena exhibit both inductive as well as deductive aspects. The processes of induction and deduction have been studied systematically though separately in the frameworks of computational learning and computational logic. Since cognition appears to incorporate these processes in combination, a single semantics is required within which the two can be discussed simultaneously. Robust logics are designed to address this question. They are based on the view that a knowledge-base can be made robust only if each assertion in it is verifiable empirically against and learnable from real world observations. The challenge then is to reconcile this with the advantages offered by conventional logics, in particular a sound basis for deduction. Robust logics are designed to bridge this gap while retaining computational feasibility. In this framework both the computational work as well as the accuracy of both learning and deduction are polynomially controlled. This therefore promises a computationally feasible approach to some of the traditional problems of artificial intelligence and seeks to avoid the main sources of brittleness encountered by logic-based or other programmed systems.
Automatic MiSe System Software Version 1.4803M | admin login