Abstract

In this paper, we introduce a deep RBF neural network for medical classification. The proposed classifier consists of two parts: an auto-encoder and an RBF neural network. The auto-encoder is used to decrease the number of the characteristics of the presented samples. Then, the obtained new features are presented to the RBF neural network. The design of the RBF neural networks is performed in two stages. First, the subtractive clustering method is used to define the centers of RBFs. Second, the genetic algorithm is used to optimize the widths of RBFs. To assess the proposed classifier, we perform tests over three medical datasets from the UCI machine-learning repository, and we compare its performances with other methods.

Information

Item Type:	Conference
Divisions:	» Faculty of Science and Technology
ePrint ID:	4423
Date Deposited:	2023-09-15
Further Information:	Google Scholar
URI:	https://old.univ-soukahras.dz/en/publication/article/4423

BibTex