Learning shape descriptions
Jonathan H. Connell, Michael Brady · International Joint Conference on Artificial Intelligence · 1985
We report on initial experiments with an implemented learning system whose inputs are images of two-dimensional shapes. The system first builds semantic network shape descriptions based on Brady's smoothed local symmetry representation. It learns shape models from them using a modified version of Winston's ANALOGY program. The learning program uses only positive examples, and is capable of learning disjunctive concepts. We discuss the lcarnability of shape descriptions.