Hysteresis response neural network and its applications

We give a novel neuron model whose activation function have two independent variables, one is the conventional input, and the other is its change rate. The response of the neuron has the hysteretic characteristics. And the output of the neuron is related to not only the current input but also the history input, which makes the neuron have stronger memory ability. The neural network composed by the neurons adopts the chaos optimization as the learning algorithm, which makes the network possess the ability of escaping from the local minimized point in the training process. The neural network exhibits strong ability for information processing. Good performances are shown in many applications, such as, hysteresis system modeling and the time series prediction of river´s runoff. Simulation results proved the method valid.