A partially connected neural network-based approach with application to breast cancer detection and recurrence

The fully connected feed-forward neural networks are commonly used in almost all neural networks applications, since such architecture provides the best generalisation power. However, they need large computing resources and have low speed when they are applied to large databases. The aim of this paper is to assess the effectiveness of an alternative approach, based on a partially connected neural network, using four significantly different breast cancer datasets for comparison. Thus, reducing the computing resource consumption during the classification process, and increasing the speed as well, this simplified neural network type succeeded in obtaining very good accuracy in comparison with a fully connected neural network.