D. H. Wang, Elizabeth Chang, Tharam S. Dillon
The paper explores the process of learning examples into neural nets and then extracting the information from these in the form of symbolic rules and concept hierarchies. Specifically the paper will review approach as using neural nets with both supervised and unsupervised learning to derive symbolic knowledge in the form of conjunctive rules, for problems with discrete features. It will then propose methods for: 1. treating disjunctive rules for discrete features; 2. treating problems with continuous features; 3. consider the problem of rules which represent data with low frequency of occurrence but high conceptual significance. Keywords: data mining, neural networks, supervised/unsupervised neural nets
© 2001-2024 Fundación Dialnet · Todos los derechos reservados