Data processing method and device for convolutional neural network

A data processing device and convolutional neural network technology, applied in the field of neural networks, can solve the problems of increasing data bandwidth, reducing the data processing capability of a convolutional neural network processing system, and increasing storage space, so as to reduce data bandwidth or storage space. requirements, improve data processing capabilities, and avoid the effect of repeated reading

Active Publication Date: 2018-02-27
TENCENT TECH (SHENZHEN) CO LTD
View PDF5 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, during the data expansion process of the convolutional neural network, some data will be repeatedly read many times, which will easily cause an increase in the data bandwidth or storage space required for the convolution operation, and reduce the performance of the convolutional neural network processing system. Data processing capability

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Data processing method and device for convolutional neural network
  • Data processing method and device for convolutional neural network
  • Data processing method and device for convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] This embodiment will describe the perspective of the data processing device of the convolutional neural network. The data processing device can specifically be integrated in a processor, and the processor can be a CPU, FPGA (Field Programmable Gate Array, Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit, application specific integrated circuit), GPU (Graphics Processing Unit, graphics processing unit) or coprocessor.

[0037] A data processing method of a convolutional neural network, which obtains matrix parameters of a feature matrix, and then reads corresponding data in an image data matrix according to the matrix parameters to obtain a data matrix to be expanded, and performs data processing on the data matrix to be expanded according to the matrix parameters Expanding, obtaining the expanded data, reading the corresponding amount of unexpanded data in the image data matrix, updating the unexpanded data matrix according to the unexpanded ...

Embodiment 2

[0118] According to the method described in Embodiment 1, an example will be given below for further detailed description.

[0119] In this embodiment, the data processing device of the convolutional neural network will be integrated with a coprocessor to Figure 5 The system architecture shown is illustrated as an example. The coprocessor can be an FPGA, ASIC, or other type of coprocessor.

[0120] In this embodiment, the image data matrix is ​​stored in the DDR memory of the processing system.

[0121] Such as Figure 8a As shown, a data processing method of a convolutional neural network, the specific process can be as follows:

[0122] 201. The coprocessor acquires a system parameter, where the system parameter includes a matrix parameter of a characteristic matrix.

[0123] The matrix parameter may include the number of rows and columns of the feature matrix. In this embodiment, the system parameters may also include the number of rows and columns of the graphic data m...

Embodiment 3

[0147] In order to better implement the above method, an embodiment of the present invention also provides a data processing device for a convolutional neural network, such as Figure 9a As shown, the data processing device of the convolutional neural network may include an acquisition unit 301, a reading unit 302, a data expansion unit 303 and an update unit 304, as follows:

[0148] (1) acquisition unit 301;

[0149] The obtaining unit 301 is configured to obtain matrix parameters of the feature matrix.

[0150] The feature matrix is ​​the convolution kernel of the convolution operation, also called the weight matrix, and the feature matrix can be set according to actual needs. Wherein, the matrix parameter of the feature matrix may include the number of rows and columns of the matrix, which may be called the size of the convolution kernel.

[0151] (2) reading unit 302;

[0152] The reading unit 302 is configured to read corresponding data in the image data matrix accord...

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

No PUM Login to view more

Abstract

The embodiment of the invention discloses a data processing method and device for a convolutional neural network. The embodiment of the method comprises the following steps that: obtaining the matrixparameter of a characteristic matrix; then, according to the matrix parameter, reading corresponding data in an image data matrix to obtain a data matrix to be expanded; according to the matrix parameter, carrying out data expansion on the matrix to be expanded to obtain expanded data; reading a corresponding quantity of unexpanded data in the image data matrix; and according to the unexpanded data, updating the data matrix to be expanded, and returning to execute the step of expanding the data matrix to be expanded according to the matrix parameter. By use of the scheme, the data processing ability of a convolutional neural network processing system can be improved.

Description

technical field [0001] The invention relates to the technical field of neural networks, in particular to a data processing method and device for a convolutional neural network. Background technique [0002] Neural networks and deep learning algorithms have been very successfully applied and are in the process of rapid development. The industry generally expects this new computing method to help realize more common and more complex intelligent applications. [0003] Among them, Convolutional Neural Network (CNN) plays an important role in deep learning because of its outstanding effect in the image field, and is one of the most widely used neural networks. [0004] The convolution operation of the convolutional neural network is mainly concentrated in the convolution layer, and the convolution operation of the convolutional neural network can be divided into two processes: data expansion and matrix multiplication. However, during the data expansion process of the convolutio...

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
no application Login to view more
IPC IPC(8): G06N3/02G06V10/764
CPCG06N3/02G06N3/063G06V10/94G06V10/454G06V10/764G06N3/045G06F17/16G06N3/08G06N3/04G06F18/217G06F18/24143
Inventor 张阳明高剑林章恒
Owner TENCENT TECH (SHENZHEN) CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap