On a neural network model based on non-associative learning mechanism and its application

Habituation is non-associative learning mechanism of biological neurons. This paper studied the simplified description of associative learning mechanism, and based on the classical M-P (McCulloch — Pitts) neuron model, put forward study neurons model with the ability of habituation learning, including habituation neurons. At the same time, in this paper, based on the simplified description of Learning neurons, the mathematical model of habituation neurons is designed, and habituation neurons are applied to deep convolution neural networks. It has been verified by experiment that habituation neurons have typical habituation learning ability, and can optimize the performance of convolution networks.