Convolution Neural Network (CNN) structure improving method

Abstract

The invention relates to the in-depth learning field, and especially relates to a Convolution Neural Network (CNN) structure improving method; the method comprises the following steps: a, using a fractional max pooling (FMP) principle to change a maximum value pooling layer in a conventional CNN structure into fractional orders, thus realizing sampling dimension reduction under a random dimension; b, ensuring a shallow network structure, continuously widening the network structure, combining the fractional order maximum value pooling layer, and thus improving the network performance. The method uses the fractional order maximum value pooling principle to ensure the layer at a shallow level, can continuously widens the network structure, thus preventing a deep network to have gradient vanish and weight failure phenomenon in a training process, and causing the CNN structure hard to train. The method can realize equal or better performance with the deep CNN structure, and uses less network parameters, thus providing obvious performance advantages.

Description

technical field [0001] The invention relates to the field of deep learning, and relates to a method based on the principle of fractional maximum pooling combined with the means of widening the shallow network to improve the performance of the convolutional neural network. In particular, a method for improving the structure of convolutional neural networks. Background technique [0002] Convolutional Neural Network (Convolutional Neural Network) is the most popular data-driven method in the field of pattern recognition recently. It has been proved by a large number of scientific research experiments that it has high performance in the field of image classification tasks. Such as figure 1 As shown, the traditional convolutional neural network structure is composed of the following parts: (1) input layer; (2) convolutional layer; (3) maximum pooling layer; (4) fully connected layer. The convolutional layer contains multiple convolutional operators acting on the image of the i...

AI Technical Summary

Problems solved by technology

[0006] The problem addressed by the present invention is that the gradient of the training process disappears due to the deepening of the network structure, resulting in low training efficiency. A shallow and wide convolutional neural network structure based on fractional maximum pooling is proposed to avoid the above problems. While improving the performance of the network

Images