A new ANN model and factor analysis of its optimization parameters

A model is presented in which the neuron plays a role in the learning phase through its adaptation of two parameters: the temperature and the level of connectivity of a given neuron. This method permits a neuron to adjust its own local temperature and the local activity of the connections. An extended back-propagation algorithm that includes both synaptical and neuronal parameters is applied to a logic type of problem, specifically, the Exclusive-OR (XOR). This results in reducing the number of cycles of the learning phase.<
>