Neural networks for breast cancer diagnosis

Abstract

Breast cancer diagnosis has been approached by various machine learning techniques for many years. The paper describes two neural network based approaches to breast cancer diagnosis, both of which have displayed good generalisation. The first approach is based on evolutionary artificial neural networks. In this approach, a feedforward neural network is evolved using an evolutionary programming algorithm. Both the weights and architectures (i.e., connectivity of the network) are evolved in the same evolutionary process. The network may grow as well as shrink. The second approach is based on neural network ensembles. In this approach, a number of feedforward neural networks are trained simultaneously in order to solve the breast cancer diagnosis problem cooperatively. The basic idea behind using a group of neural networks rather than a monolithic one is divide-and-conquer. The negative correlation training algorithm we used attempts to decompose a problem automatically and then solve them. We illustrate how negative correlation helps a group of neural networks learn using a real world time series prediction problem.