The Function Representation of Artificial Neural Network

In this paper, the artificial neural network (ANN) is functio nalized utilizing the activation integral, so its principle and design can be easily deduced. In the first part, the linear activation function is abstracted from normal activ ation functions, and the activation integral in continuous form is established. Then the discre te form of activation integral of composite function and multivariate function are introduc ed. Finally, the standard discrete activation integral of any function is derived, which is the discrete activation integral with linear activation function integrated. In the second part, the activation integral theory is introduced into ANN, and the main network structure types ar e analyzed, and the corresponding function forms are obtained.