Third-order generalization and a new approach to systematically categorizing higher-order generalization
R.S. Neville · Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005. · 2006
Higher-order generalization is a means of categorizing different types of generalization. The paper presents a framework within which higher-order generalization can be evaluated in a detailed and systematic way. Previous research divided generalization into three categories. However, these categories were fuzzy and imprecise. This paper further refines existing definitions by first assigning each category a logical predicate that it must fulfil in order to achieve a specific order (type) of generalization. Then, it breaks the orders down into four different categories in a detailed and systematic way. The paper focuses on early (initial) results; some of the aims have been demonstrated and amplified through the experimental work.