Maulsby, D. & Witten, I.H. (1995). Learning to describe data in actions. In Proceedings of Workshop on Programming by Demonstration, Twelfth International Conference on Machine Learning, Lake Tahoe, USA, July 9th 1995 (pp. 65-73).
Abstract
Traditional machine learning algorithms have failed to serve the needs of systems for Programming by Demonstration (PBD), which require interaction with a user (a teacher) and a task environment. We argue that traditional learning algorithms fail for two reasons: they do not cope with the ambiguous instructions that users provide in addition to examples; and their learning criterion requires only that concepts classify examples to some degree of accuracy, ignoring the other ways in which an active agent might use concepts. We show how a classic concept learning algorithm can be adapted for use in PBD by replacing the learning criterion with a set of instructional and utility criteria, and by replacing a statistical preference bias with a set of heuristics that exploit user hints and background knowledge to focus attention.
Rights
This article has been published in Proceedings of Twelfth International Conference on Machine Learning, Lake Tahoe, USA, July 9th 1995. © 1995 the authors.
