An Analytical Framework for Learning Systems
Robert C. Holte · 1988
The problem addresses in this thesis is that of defining a set of concepts and techniques that facilitate the comparison and analysis of learning systems. Systems are modelled in terms of certain abstract processes and bodies of information. Different types of systems correspond to different ways of representing the model. Systems of different types are compared using behavior-preserving transformations. Formal definitions are given for representation and generative structure of a system. These and related concepts, such as bias and implicit knowledge, facilitate the analysis of a system''s efficiency and its use of task-specific knowledge.