Evolutionary Neural Logic Networks in Two Medical Decision Tasks.

Abstract

Two real-world problems of the medical domain are addressed in this work using a novel approach belonging to the area of neural-symbolic systems. Specifically,we apply evolutionary techniques for the development of neural logic networks of arbitrary length and topology. The evolutionary algorithm is consisted of grammar guided genetic programming using cellular encoding for the representation of neural logic networks into population individuals. The application area is consisted of the diagnosis of patient postoperative treatment and the diagnosis of the Breast cancer. The extracted solutions maintain their interpretability into simple and comprehensible logical rules. The overall system is shown capable to generate arbitrarily connected and interpretable evolved solutions leading to potential knowledge extraction.