Improved Cerebellar Model Neural Network and Its Application in Time Series Prediction

To tackle the key problem of the conflict between generalization ability and storage space capacity in the cerebellar model neural network (CMNN), an improved cerebellar model neural network-fuzzy membership cerebellar model neural network (FM-CMNN) was proposed to solve the time series prediction problem of nonlinear dynamic systems. First, FM-CMNN, while retaining the address mapping mode of the original CMNN input variables, the bell fuzzy membership function was introduced into the CMNN storage space, so as to improve the generalization ability of the network without increasing the quantization series. Then, the gradient descent algorithm was used to update the weights of the network to improve the approximation strength of the network. Finally, the reliability of the network was verified by the nonlinear time series prediction benchmark experiment and the water quality parameter prediction experiment in sewage treatment.