Method and apparatus for learning and applying multilayer neural network model, and storage medium

摘要

PROBLEM TO BE SOLVED: To achieve a better balance between simplifying a multilayer neural network model and maintaining performance. A multilayer neural network model uses, for a convolutional layer, a step of expanding the number of channels of a filter of the convolutional layer and a filter after expanding the number of channels based on data corresponding to application requirements during forward propagation. And the step of performing data calculation of the convolutional layer, and the step of outputting the application result after completing the forward propagation. [Selection diagram] Fig. 7