ICHIKI@SRI-STRIPE.ARPA (Joani Ichiki) (10/07/86)
L. Saitta (Dipartimento di Informatica, Universita di Torino, Italy) will present his talk entitled, "AUTOMATED CONCEPT ACQUISITION IN NOISY ENVIRONMENTS," 10/7/86 in EK242 at 11:00am. Abstract follows. This paper presents a system which performs automated concept acquisition from examples and has been especially designed to work in errorful and noisy environments. The adopted learning methodology is aimed to the target problem of finding discriminant descriptions of a given set of concepts and both examples and counterexamples are used. The learning knowledge is expressed in the form of production rules, organized into separate clusters, linked together in a graph structure; the condition part of the rules, corresponding to descriptions of relevant aspects of the concepts, is expressed by means of a first order logic based language, enriched with constructs suitable to handle uncertainty and vagueness and to increase readability by a human user. A continuous-valued semantics is associated to this language and each rule is affected by a certainty factor. Learning is considered as a cyclic process of knowledge extraction, validation and refinement; the control of the cycle is left to the teacher. Knowledge extraction proceeds through a process of specialization, rather than generalization, and utilizes a technique of problem reduction to contain the computational complexity. Moreover, the search strategy is strongly focalized by means of task-oriented but domain-independent heuristics, trying to emulate the learning mechanism of a human being, faced to find discrimination rules from a set of examples. Several criteria are proposed for evaluating the acquired knowledge; these criteria are used to guide the process of knowledge refinement. The methodology has been tested on a problem in the field of speech recognition and the obtained experimental results are reported and discussed. -------