clarke@utcsri.UUCP (05/01/87)
(Does not include the two "late" announcements posted earlier (!).) SUMMARY: THEORY SEMINAR, Monday, May 4, 3 pm, GB120 -- Lajos Ronyai: "Complexity of Matrix Algebras" A.I. SEMINAR, Tuesday, May 5, 3 pm, GB120 -- Tom M. Mitchell: "Toward a Learning Robot" THEORY SEMINAR, Thursday, May 7, 11 am, GB120 -- Jeffrey B. Sidney: "Geometric Containment and Partial Orders" A.I. SEMINAR, Friday, May 8, 11 am, SF1105 -- Brian Cantwell Smith: "Designing a Situated Inference Engine" ---------------- (SF = Sandford Fleming Building, 10 King's College Road) (GB = Galbraith Building, 35 St. George Street) THEORY SEMINAR, Monday, May 4, 3 pm, GB120 Dr. Lajos Ronyai University of Chicago ``Complexity of Matrix Algebras" The problem of factoring polynomials over a field is usually solved in two main steps. First we remove the multiple factors and then find the irreducible factors of the remaining squarefree polynomial. This decompo- sition can naturally be generalized to matrix algebras, i.e. is a linear space of matrices closed under matrix multiplication. The radical corresponds to the multiple factors; and the so-called minimal ideals of the radical-free part to the irreducible factors. We discuss the question of finding the structure (radical and minimal ideals) of matrix algebras over finite fields and algebraic number fields. The key problem is to find out if the algebra contains any singular matrices (other than zero). We solve this problem in Las Vegas polynomial time over finite fields. We indicate that the same problem over the rationals might be substantially more difficult. Indeed, under a number theoretic hypothesis, we show that hard number theoretic problems admit a randomized polynomial time reduction to this problem. A.I. SEMINAR, Tuesday, May 5, 3 pm, GB120 Professor Tom M. Mitchell Carnegie Mellon University ``Toward a Learning Robot" To improve its performance within its world, a robot will have to learn two kinds of knowledge: (1) a model of its world: the effects of its actions on the world and the effects of the world on it, and (2) effi- cient problem solving strategies relative to this model of the world. We have recently begun a research project to develop a learning hand-eye system whose goal is to improve its model of its world (i.e., its table) as well as the efficacy of its problem solving strategies. This talk will present our present, tentative, explorations toward such a learn- ing robot. THEORY SEMINAR, Thursday, May 7, 11 am, GB120 Professor Jeffrey B. Sidney University of Ottawa ``Geometric Containment and Partial Orders" The following question will be addressed: Given a class of geometric figures, how many real variables are required to parameterize the class in such a fashion that one figure from the class is contained in another (perhaps after translation, rotation and even reflection) if and only if the parameter values for the first figure are no greater than those for the second figure? In particular, will finitely many parameters suffice? For rectangles it is known that finite parameterizations will not suf- fice. This result is generalized in an abstract theorem, which is applied to other classes of geometric objects (right circular cylinders, isosceles triangles) and to a certain natural algebraic partially ordered set. Other implications of the abstract theorem are also discussed. A.I. SEMINAR, Friday, May 8, 11 am, SF1105 Professor Brian Cantwell Smith Palo Alto Research Center "Designing a Situated Inference Engine" According to the "situated" perspective, the structure of both language and reasoning arise out of mutual constraints between a system and its embedding environment. The Situated Inference Engine is a pilot pro- ject exploring this perspective in the context of a simple scheduling sys- tem. In this talk I will review the situated perspective, introduce the SIE, present the language it will use to communicate with its users, and discuss our current conception of its internal architecture. -- Jim Clarke -- Dept. of Computer Science, Univ. of Toronto, Canada M5S 1A4 (416) 978-4058 {allegra,cornell,decvax,linus,utzoo}!utcsri!clarke