An ensemble average classifier for pattern recognition machines

Abstract The ensemble average classification method introduced here is a new nonparametric classification procedure. In this method, ensemble average of training pattern vectors in each class is stored in a computer memory. Classification of an unknown pattern vector depends primarily on the difference between the stored ensemble average vectors and the unknown pattern vector. Performance of the new method in comparison with Bayesian (optimal) and perceptron classifiers has been studied through a series of computer experiments. The results obtained showed that the new method provides as higher classification rates as the Bayes classifier. However, it requires less computation complexity and higher storage memory than both Bayesian and perceptron classifiers.