Convolutional neural network accelerator based on Winograd sparse algorithm

A convolutional neural network and accelerator technology, applied in the field of convolutional neural network accelerators, can solve problems such as high computational complexity, and achieve the effects of good acceleration effect, reduced pressure, and simple structure

Pending Publication Date: 2020-02-18
NAT UNIV OF DEFENSE TECH
View PDF2 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Judging from the current trend, the topology of the convolutional neural network continues to deepen, which brings higher computational complexity.

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
  • Convolutional neural network accelerator based on Winograd sparse algorithm
  • Convolutional neural network accelerator based on Winograd sparse algorithm
  • Convolutional neural network accelerator based on Winograd sparse algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0034] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0035] Such as figure 1 Shown, a kind of convolutional neural network accelerator based on Winograd sparse algorithm of the present invention, it comprises:

[0036] The control module (topcontrol) is responsible for the movement of data;

[0037] Buffer module (buffer), used for temporary storage of load data,

[0038] Operational modules (PUs) are responsible for completing the operation of the Winograd sparse algorithm.

[0039] In the read phase, the address is sent by the control module (top-level control), and the input buffer and weight buffer read the data in the external DRAM;

[0040] In the data operation stage, the operation module reads the input data, weight data and weight index from the buffer module to complete the convolution operation;

[0041] In the sending phase, when the output completes the final accumulati...

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 discloses a convolutional neural network accelerator based on a Winograd sparse algorithm. The convolutional neural network accelerator comprises; a control module which is used for moving data; a buffer module buffer which is used for temporarily storing load data, and an operation module which is used for completing operation of a Winograd sparse algorithm. In the reading stage, the control module sends an address, an input cache and a weight cache to read data in the external DRAM; in the data operation stage, the operation module reads input data, weight data and a weight index from a buffer module to complete convolution operation; and in the sending stage, when the output finishes the final accumulation operation, the data is sent to the external DRAM through the outputcache, and finally calculation is finished. The convolutional neural network accelerator has the advantages of simple structure, easiness in implementation, good acceleration effect and the like.

Description

technical field [0001] The present invention mainly relates to the technical field of convolutional neural network, in particular to a convolutional neural network accelerator based on Winograd sparse algorithm. Background technique [0002] Convolutional neural networks are currently widely used in various computer fields, such as image recognition, recommendation systems, and language processing. However, the time to train and derive a convolutional neural network is unbearable. The reason is that the convolutional neural network introduces a convolutional layer, which increases the computational complexity in the network and brings a huge workload. This is difficult for current CPUs or embedded processors to solve. [0003] In order to solve this problem, many solutions have been proposed, such as using GPU acceleration during the operation, or using FPGA, custom ASIC and other hardware to complete the convolution operation. Most of these schemes use the parallelism of...

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 郭阳徐睿马胜刘胜陈海燕王耀华
Owner NAT UNIV OF DEFENSE TECH
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