BUILDING A RESONANCE FRAMEWORK FOR WORD RECOGNITION USING DESIGN AND SYSTEM PRINCIPLES

Basic elements of a principle-based approach to model development are presented. The approach is needed to understand the operation of models capable of complex behavior. The use of principles facilitatates both assignment of explanatory credit and blame when testing models and guides refinement of models when they fail. Two types of principles are distinguished: design and system. Design principles relate model behavior to observable human behavior. System principles relate model behavior to assumptions about a model's formal structure (architectural axioms). The use of, and relationship between, such principles is illustrated by building the theoretical framework of resonance (S. Grossberg and G. O. Stone, 1986) through the successive addition of principles.