We have modeled the neuronal circuit that conveys mechanosensory input onto a pair of serotonergic neurons in the nervous system of the leech. The objective of this work is to use an artificial neural networks (ANN) in the investigation of the interneuronal layer of this circuit, which represents an unknown population of cells. The ANN is a three layer perceptron, whose input units represent the mechanosensory neurons, the output unit represents the serotonergic neurons and the hidden units were represented as a saturation function rather than as linear one. The weights and thresholds were adjusted using the backpropagation algorithm. The ANN parameters were correlated with specific biological parameters enabling to test and classify the resulting configurations according to physiological and experimental considerations. We obtained a finite and restricted number of solutions that can be experimentally tested.
展开▼