New Neural Networks Based on Taylor Series and their Research

摘要

This paper is Mining in the essence of neural networks, and constructing 4 types of neural networks: (1) to construct a neural network based on Taylor series; (2) to construct a Taylor component neural network which brings in a Radial Basis Function neuron as a prefix; (3) to construct a Fourier component neural network Because of the relationships between these functions, the Taylor component NN and the Fourier component NN can be called Gauss series NN equivalently; (4) to construct a Gauss series Clustering neural network and to prove its equivalence with RBF NN in a limit situation. The development of new types of neural Networks is playing an important role either to promote deepening study of neural networks theory or to provide new methods for applications.