Optimal Multilayer Perceptron Research Articles

Classification of the strain and growth phase of cyanobacteria in potable water using an electronic nose system

An electronic nose comprising an array of six commercial odour sensors has been used to monitor not only different strains, but also the growth phase, of cyanobacteria which is normally called blue green algal. A series of experiments were carried out to analyse the nature of two closely related strains of cyanobacteria, Microcystis aeruginosa PCC 7806 that produces a toxin and PCC 7941 that does not. The authors have constructed a measurement system for the testing of the cyanobacteria in water over a period of up to 40 days. After some pre-processing to remove the variation associated with running the electronic nose in ambient air, the two different strains, and their growth phase, with principal component, were classified analysis, multilayer perceptron (MLP), learning vector quantisation (LVQ), and fuzzy ARTMAP. The optimal MLP network was found to classify correctly 97.1% of unknown non-toxic and 100% of unkhown toxic cyanobacteria. The optimal LVQ and fuzzy ARTMAP algorithms were able to classify 100% of both strains of cyanobacteria. The accuracy of MLP, LVQ and fuzzy ARTMAP algorithms with the four different growth phases of toxic cyanobacteria was 92.3%, 95.1% and 92.3%, respectively.

Real-time parallel processing implementation for word recognition using MLP and HMM

This paper presents a relevant implementation of the completeisolated word recognition system, oriented toward a multiprocessor solution as a vehicle to fulfil two important requirements: flexibility, necessary for investigation into the optimum multi-layer perceptron and hidden Markov model configuration for the solution of each specific problem; and real-time processing capability, in order to convert the optimum configuration obtained through simulation into an effective real-time speech recognition system.

An optimum multilayer perceptron neural receiver for signal detection

The M-input optimum likelihood-ratio receiver is generalized by considering the case of different signal amplitudes on the receiver primary input lines. Using the more general likelihood-ratio receiver as a reference, an equivalent optimum multilayer perceptron neural network (or neural receiver) is identified for detecting the presence of an M-dimensional target signal corrupted by bandlimited white Gaussian noise. Analytical results are supported by Monte Carlo simulation runs which indicate that the detection capability of the proposed neural receiver is not sensitive to the level of training or number of patterns in the training set.