A computational model of the olfactory bulb

In this paper the authors propose a model of the olfactory bulb, which is able to replicate as accurately as possible the architecture and processing methods of the natural olfactory system, and is also formulated in computational terms. The development of the model is based on the assumption that in the olfactory bulb learning occurs and that the network formed by the neurons belonging to it is able to discriminate odours very selectively. The model was implemented by means of an artificial neural network whose architecture replicates the structure of the bulbar neuron pool. In order to evaluate the applicability of the network in an artificial nose, the behaviour of the neural net was evaluated on the basis of its ability to classify correctly experimental data from an array of scarcely selective conducting polymer sensors concerning five odour classes.