Evaluating the Performance of a Centroid-Based Probabilistic Neural Network

In this article is proposed a technique which uses centroids together with Probabilistic Neural Network to minimize some disadvantages of this net, such as the storage space for the neural network weights and linear time complexity order with the number of training samples. In the experiments carry out the memory usage and classification time were drastically reduced. Besides, the quality of the results was also considering improved by the a priory probability, when using it with theses centroids.