Convolutional neural network construction method

Abstract

The present invention discloses a convolutional neural network construction method, and belongs to the neural network technology field. According to the present invention, when the convolutional neural network transfers forward, on each original convolution kernel, and by manually adjusting the dot product of a kernel and the original convolution kernel, the original convolution kernel is modulated to obtain the modulated convolution kernel, and the modulated convolution kernel is used to substitute for the original convolution kernel to transfer the neural network forward, thereby realizing afeature enhancement effect. According to the method of the present invention, the neural network is optimized greatly, so that the total number of the kernels that the network must learn is reduced.In addition, by modulating to generate the sub-convolution kernels to arrange the kernels of the redundant learning in an original network structure, a model compression purpose also can be achieved.

Description

technical field [0001] The invention relates to the technical fields of image recognition, artificial intelligence and neural network, in particular to a method for constructing a convolutional neural network. Background technique [0002] In recent years, with the emergence of ultra-large-scale classified data sets and parallel computing tools GPU, deep convolutional neural networks (deep convolution neural networks, DCNNs) have developed rapidly in the field of computer vision, and have received extensive attention from the academic community. This end-to-end network drives training through a large number of data training samples, and independently learns model parameters with the help of optimization algorithms such as stochastic gradient descent. A breakthrough has been made. [0003] The improvement of DCNNs performance depends on the expansion of training data and complex model structure. However, many practical problems in real life are usually only supported by smal...

AI Technical Summary

Problems solved by technology

In addition, traditional DCNNs usually have redundant learning convolution kernels during the training process. When the number of layers of the neural network increases, the data of the network will increase rapidly, so after training, the saved model will also take up a lot of storage space.

[0004] It can be seen that the deep convolutional neural network in the prior art has the problems of large training samples and consuming storage space

Images