Modelling of the qualitative discrimination of odours in the first two layers of olfactory system by Jutten and Hérault algorithm.

The olfactory system discriminates many different odours although membrane receptors, and neuroreceptors that bear them, are only broadly tuned to volatile molecules. We investigated how mixed and redundant information carried by neuroreceptors could be efficiently coded by output neurons in the second neuronal layer in the brain. We showed that a good discrimination is yielded if neuroreceptors that terminate on the same output neuron belong to several types, if these types are in different proportions for each output neuron, and if each output neuron exerts an action on all others so that their independence is maximized, according to the Jutten and Hérault algorithm for separating linear mixtures of independent sources. The proposed model details the computations made by neurons and leads to a better understanding of several observed properties of the olfactory system, among which the inhibitory actions of local neurons of the second layer and the identifiability of insect glomeruli.