Title: A two-phase genetic learning of a neural classifier application in medical diagnostic

Authors: Mansouria Sekkal; Mohammed Amine Chikh

Addresses: Laboratory of Biomedical Engineering, Department of Electrical Engineering, Tlemcen University, Tlemcen 13000, Algeria ' Laboratory of Biomedical Engineering, Department of Electrical Engineering, Tlemcen University, Tlemcen 13000, Algeria

Abstract: In this paper, we propose a procedure to choose an initial population at the beginning of an evolutionary process in neural networks. In the first phase, we take N examples of the learning base (N represents size of initial pop for second phase evaluation) and find the best classifier synaptic weights for each individual example using a Neuro-Genetic Classifier (NGC). In the second phase, we use a global genetic learning database, as the initial population is represent by all final weights of the first learning phase. The objective of this method is to ameliorate the performance of NGCs with a lower computational cost. The results show that our proposal considerably improved the efficiency of previous approaches. We use several medical databases to validate our results.

Keywords: artificial neural networks; genetic algorithm; classification; choice of initial population.

DOI: 10.1504/IJBET.2018.094427

International Journal of Biomedical Engineering and Technology, 2018 Vol.28 No.1, pp.38 - 52

Received: 16 Feb 2016
Accepted: 26 Aug 2016
Published online: 03 Sep 2018 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article