INDUCING FUNCTIONS IN ROBOT DOMAINS
David Pauli, Bruce MacDonald · Libraries and Cultural Resources (University of Calgary) · 1988
This paper describes a reimplementation of a function induction algorithm that is part of Peter Andreae's robot procedure learning system NODDY. Results are given, the implementation is compared to Andreae's version, and the BACON and COPER systems are compared to both. The discussion focuses on the representation of knowledge--in particular on the representation of inverse operators and the induced expression--and the strength of argument typing.