Parallel-series perceptrons for the simultaneous determination of odor classes and concentrations

Abstract

The simultaneous determination of odor classes and concentrations is solved by a kind of parallel-series perceptron models. Two groups of parallel single-output perceptrons are in series, and the former is responsible for classification, and the latter for location. The number of parallel perceptrons is equal to the number of odor classes. A multi-class learning problem is first decomposed into multiple two-class problems, and then solved by multiple parallel perceptrons, one by one. Each training subset is composed of the most necessary samples. And furthermore, some virtual samples are added to the weak side of any two-class learning subsets in order to arrive at a virtual balance. The experimental results for 4 kinds of fragrant materials show that the proposed parallel-series perceptrons with the electronic nose are effective. © Springer-Verlag Berlin Heidelberg 2007.