Adaptive resonance neural classifier for identification of gases/odours using an integrated sensor array

Drift counteraction with multiple self-organising maps for an electronic nose

“…In case of not performing filtering operations the network shows a good performance only to small threshold value (b < 10 −4 ). From this figure, it can pick out the best mSom performances (or the corresponding optimal threshold values) and their results are compared against previously version of mSom (with the constant refresh rate fixed to 50) 4 and another techniques as RBF and fuzzy ARTMAP in terms of errors committed in recognising the given data set with the same training and testing sets (as shown in Table 3). As regards these chosen mSom performances, Figs.…”

“…The previous version used a fixed refresh rate parameter to start autonomous retraining process to compensate the varying probability distribution of data due to drift. The new version 4 The constant refresh rate defines the number of unknown input vectors before of starting a new retraining process.…”

“…At the moment there is an increasing interest in the methods based on self-organising neural networks [5,6] and adaptive resonance theory for their ability of processing non-stationary sources of data such as sensor output subjected to drift effects [3,4].…”

A Radial Basis Function Neural Network Classifier for the Discrimination of Individual Odor Using Responses of Thick-Film Tin-Oxide Sensors

“…This has led to the shifting of the interest of researchers towards artificial neural network (ANN) techniques which apart from being massively parallel in nature, are also capable of handling nonlinear transduction properties. BP trained ANN, genetically trained ANN [14], radial basis function neural networks [15], adaptive resonance theory [16], self-organizing networks [17] and the combination of fuzzy concept and neural networks [18] have been applied to odor/gas classification problems. The success of odor/gas identification task not only depends upon the nature of the classifier but also upon the nature of the input data set [19].…”

A fuzzy logic based neural network classifier for qualitative classification of odors/gases

“…This has led to the shifting of the interest of researchers towards Artificial Neural Network (ANN) techniques which apart from being massively parallel in nature, are also capable of handling non-linear transduction properties. Backpropagation trained ANN, genetically trained ANN [14], radial basis function neural networks [15], adaptive resonance theory [16], and self-organizing networks [17] have been applied to odor/gas classification problems. This course of development has also led to the emergence of hybrid techniques in which some other pattern recognition technique is used along with the ANN, either as different stages of a system, or as a single system inculcating the salient features of more than one technique.…”