| Fawcett, T. E. (1990). Feature discovery for inductive concept learning, (Coins Technical Report 90-15), Amherst, MA: University of Massachusetts, Department of Computer and Information Science. |
....is effective in this domain, the results will be of great practical significance. The theory will also be applied to the game of Othello. 11 This problem was chosen because move selection in Othello is an intractable search problem, and because two performance programs, 12 a learning program [Fawcett, 1990], and a catalog of sophisticated features [Mitchell, 1984] are all available for comparison. The theory will be applied to at least one other problem. This will provide a set of five domains in which to test the theory. Two of the domains are classic constraint satisfaction problems (Eight ....
Fawcett, T. E. (1990). Feature discovery for inductive concept learning, (Coins Technical Report 90-15), Amherst, MA: University of Massachusetts, Department of Computer and Information Science.
....is active, unsupervised and builds upon the modeling task (that is feature space is built by the agent) for a dynamic environment. Our work can also be considered as a discovery system where new features are discovered and used in further learning and model usage [Zhao 1994, Yip 1991, Thrun , Fawcett 1993 ] Computer Vision is often concerned with recognition of geometric shapes of physical objects [Ullman ] For us, recognition of geometrical shapes is not central, though our agent can do this. Our concern is to know the behaviors of the objects. We start with low level data such as the vision ....
Fawcett T E. 1993. Feature Discovery for Inductive Concept Learning.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC