Model of biological ANN based on homeostatic neurons
Martin Růžek, Tomáš Brandejský · 2011
This paper presents an artificial neural network that is composed of homeostatic neurons. The homeostatic neron is based on McCulloch-Pitts model of artificial neuron with some modifications. The basic requirements for this model is the simplicity and similarity to biological neuron. It is supposed that the neuron is able to maintain its parameters in a way that maximizes its chances to survive. The probability of survival in the neural network depends on how well is accepted the output by other neurons. The homeostatic neurons form a neural network, that is aimed to model biological neural network. The purpose of this network is not to perform any specific task (such as prediction) but to simulate precesses in human brain, e. g. the loss of attention.