Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Parallel optimization method of convolutional neural network

A convolutional neural network and optimization method technology, applied to biological neural network models, neural architectures, etc., can solve the problems of no convolution algorithm optimization, low parallel efficiency of convolutional neural network models, and waste of processing power. Achieve the effect of reducing time complexity, reducing computing overhead, and reducing complexity

Active Publication Date: 2019-08-23
CHENGDU UNIV OF INFORMATION TECH
View PDF4 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0009] 1. The model parallel efficiency of convolutional neural network is low; 2. There is still a lot of waste in the processing power of large-scale cluster parallelism on a single node; 3. There is no algorithmic optimization of the convolution operation in convolutional neural network. Optimization; 4. The continuity of the data is not considered in the specific calculation, and the vectorization is not used better

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
  • Parallel optimization method of convolutional neural network
  • Parallel optimization method of convolutional neural network
  • Parallel optimization method of convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0031] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in combination with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are exemplary only, and are not intended to limit the scope of the present invention. Also, in the following description, descriptions of well-known structures and techniques are omitted to avoid unnecessarily obscuring the concept of the present invention.

[0032] The parallel optimization method of the convolutional neural network of the present invention comprises:

[0033] Step 1: Replace the winograd algorithm f(4x4,3x3) with f(2x2,3x3) to reduce the time complexity and the number of multiplication operations.

[0034] Step 2: For the for loop structure part that does not have loop data dependence, use OpenMP to open up multiple threads for calculation, so as to achieve...

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 invention relates to a parallel optimization method for a convolutional neural network, and the method comprises the steps: carrying out the convolution kernel operation of the convolutional neural network through employing a winograd algorithm f (2x2, 3x3), so as to reduce the time complexity and reduce the number of times of multiplication; for the for circulation structure part without circulation data dependence, using an OpenMP for opening up a plurality of threads to carry out operation; carrying out vectorization processing on the data operation part with the same mode, so that one-time instruction multiple operation can be realized. The convolutional neural network program improved by adopting the method provided by the invention can greatly improve the parallel efficiency, reduce the operation complexity, fundamentally reduce the operation expenditure, and reduce the operation time of the program.

Description

technical field [0001] The invention relates to the technical field of neural networks, in particular to a parallel optimization method for convolutional neural networks. Background technique [0002] Convolutional neural network (CNN) is a recently developed feedback neural network. Its relatively efficient recognition rate and the need not to do too much complex preprocessing of images in the early stage make the convolutional neural network in the image mode Classification has been widely used. The basic structure of convolutional neural network (CNN) includes feature extraction layer and feature mapping layer. In the first layer, the input of each neuron is connected with the local receptive field of the previous layer to obtain the local features of this area. Then obtain the local features of other regions, as well as the positional relationship between each local feature. In the second layer, each shared feature is calculated in multiple computing layers, and each ...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04
CPCG06N3/045
Inventor 黄家豪王铁军朱旭辉魏敏陈海宁杨昊赵长名吴涛黄敏吴锡
Owner CHENGDU UNIV OF INFORMATION TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products