ARTIFICIAL NEURON WITH HOMEOSTATIC LEARNING

This article presents a new model of artificial neuron. This neuron is able to find independently its homeostatic position during the process of learning. The basic idea of the learning mode is that the neuron is trying to maximize its importance for other neurons. The learning phase is completely autonomous: the neuron does not need any superior structure that informs it about the correctness of its function. The inputs for the learning are the weights of other neurons that accept the output of the reference neuron. The neuron is able to change its parameters in the way that maximizes the weights of other connected neurons. This article presents the theoretical elementals of this concept and the results that were achieved during the tests. This neuron is the basic element of a planned network.