Convolution neural network training and testing method and training and testing device
A technology of convolutional neural network and training method, which is applied in the field of convolutional neural network training, testing method and training, and testing device, and can solve the problems of storage capacity and calculation capacity hindering popularization, unfavorable promotion and application of embedded devices, low CPU performance, etc. problem, to achieve the effect of reducing the amount of data, reducing the amount of storage, and reducing the amount of calculation
Inactive Publication Date: 2018-04-06
NEUSOFT CORP +1
View PDF0 Cites 11 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0012] However, the huge amount of storage and calculation of the convolutional neural network hinders its promotion in practical applications. For example, for embedded devices, its CPU performance is low, and the current calculation method of the convolutional neural network is not conducive to the promotion of embedded devices. application
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0071] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
[0072] On the one hand, the inventor found in research that with the development of science and technology, embedded devices are more and more commonly used in practice, and embedded devices may include mobile terminals, small robots, and vehicle-mounted terminals. For example, the functions that can be realized by the mobile terminal are also mo...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a convolution neural network training and testing method and training and testing device. According to the training method, when each convolution kernel is initialized in a first step, each convolution kernel is initialized according to a mask array and a neural network structure, and when the convolution kernel is updated in a fifth step, the initialization convolution kernel is updated according to the mask array and the neural network structure. Since the initialization convolution kernel and an update convolution kernel are obtained by masking by using the mask array, which means that the masks of a part of elements in the initialization convolution kernel and the update convolution kernel are zero. Thus, finally, the obtained storage of convolution kernel datain a convolutional neural network model is reduced. The storage of the convolutional neural network model obtained by training is reduced. When the convolutional neural network model obtained by training is used to calculate, and since the data amount of the convolution kernel data involved in the calculation is reduced, the calculation amount of using the convolutional neural network model to calculate is reduced.
Description
technical field [0001] The invention relates to the technical field of deep learning, in particular to a convolutional neural network training and testing method and training and testing device. Background technique [0002] Deep learning has shown great advantages in the fields of image detection and speech recognition, and the important algorithm used in it is the use of convolutional neural network models. [0003] The following briefly introduces the training method of the convolutional neural network model in the prior art. [0004] see figure 1 , which is a flow chart of training a convolutional neural network model in the prior art. [0005] S101: Randomly initialize the initial value data of each convolution kernel of the neural network according to a normal distribution. [0006] S102: Calculate forward according to the neural network structure. [0007] S103: If the loss function reaches a given threshold, execute S107; otherwise, execute S104. [0008] S104: ...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More IPC IPC(8): G06N3/04G06N3/08
CPCG06N3/08G06N3/045
Inventor 贾书军程帅袁淮刘威
Owner NEUSOFT CORP