Deep learning mode recognition method for vector core convolutional neural network

Abstract

The invention discloses a deep learning pattern recognition method of a vector kernel convolutional neural network, the method comprising: (1) designing the structure of the vector kernel convolutional neural network, including an input layer, a convolutional layer, and at least one fully connected layer , and the output layer of the soft-max classifier; where the input is an image of size m×n, the kernels of layer l are vectors of size pl×1 or vectors of size 1×ql, and Nl represents the size of layer l The number of convolution kernels; select the values ​​of pl (or ql) and Nl based on experience, the convolution operation step size of the l layer is sl (sl≥1), and the activation function is selected as a corrected linear unit; (2) set The number of iterations of the convolutional neural network, select the cost function, and use the training samples {(x1,y1),...,(xD,yD)} to learn the parameters of the vector kernel convolutional neural network according to the backpropagation algorithm;( 3) Judging whether the number of iterations is completed, if not, continue training; if the number of iterations is completed, input the test sample into the trained network model for testing, and obtain the test result.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and deep learning. The specific content is a deep learning pattern recognition method of vector kernel convolutional neural network. Background technique [0002] Deep learning is a new research direction in the field of artificial intelligence. In recent years, breakthroughs have been made in various applications such as image recognition and computer vision. With the continuous development of deep learning, deep neural network models such as AutoEncoder (AE), Deep Belief Networks (DBNs), and Convolutional Neural Networks (CNN) have emerged. Among them, CNN has been widely used because of its characteristics of weight sharing, local receptive field, and dimensionality reduction. However, CNN has a large amount of weight data. As the number of layers increases, the training time will be longer and the consumption of computing resources will be larger. Contents of the invention [...

AI Technical Summary

Problems solved by technology

However, CNN has a large amount of weight data. With the increase of the number of layers, the training time is longer and the consumption of computing resources is larger.

Images