Genetic Algorithm Design of Neural Net Based Electronic Nose
The training of a multi-layer perceptron using the well known back-propagation algorithm normally takes place after the neural network architecture and the initial values of various network parameters have been defined. Since the success of the training process, in terms of a fast rate of convergence and good generalisation, can be affected by the choice of the architecture and the initial network parameters, much time is spent in searching for the optimal neural paradigm. In this paper, results are presented on the use of Genetic Algorithms to determine automatically a suitable network architecture and a set of parameters from a restricted region of design space. The data-set comes from the response of the Warwick Electronic Nose to a set of simple and complex odours.
KeywordsGenetic Algorithm Hide Layer Network Parameter Binary String Electronic Nose
Unable to display preview. Download preview PDF.
- Gardner J W, Hines E L and Tang H C, ‘Detection of vapours and odours from a multisensor array using pattern recognition technique. Part 2: Artificial neural networks’, Sensors and Actuators, B, pp9-15, 1992.Google Scholar
- Rumelhart D E and McClelland J L, ‘Parallel distributed processing’, MIT press, chapter 8, 1986.Google Scholar
- Davis L, ‘Handbook of Genetic Algorithms’, Van Nostrand Reinhold, 1991.Google Scholar
- Leung C W, ‘Application of neural networks to the classification of coffee data’, MSc IT Dissertation, Department of Engineering, University of Warwick, 1991.Google Scholar
- Hines E L, Gardner J W, Fung W and Fekadu A A, ‘Improved rate of convergence in a MLP based electronic nose’, 2nd Irish Neural Network Conf., Belfast, June 1992.Google Scholar