Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Customized deep neural network model compression method and system based on cloud edge cooperation

A deep neural network and compression system technology, applied in the field of customized deep neural network model compression, can solve problems such as insufficient memory size and computing power, limited edge scenarios, and inability to deploy real-time reasoning on the edge side

Pending Publication Date: 2021-03-12
ZHEJIANG LAB
View PDF0 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The present invention can solve the problem that edge scenarios are limited by hardware constraints, such as insufficient CPU / GPU / NPU / memory size and computing power, resulting in the inability to deploy on the edge side or perform real-time reasoning; in fragmented edge scenarios, user-based Customized requirements for model compression and acceleration

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
  • Customized deep neural network model compression method and system based on cloud edge cooperation
  • Customized deep neural network model compression method and system based on cloud edge cooperation
  • Customized deep neural network model compression method and system based on cloud edge cooperation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0084] The present invention evaluates the performance of the subclass distillation algorithm based on the CIFAR-10 data set. The CIFAR-10 dataset contains 50K training data of 32×32 size and 10K test data, including 10 categories in total. Based on the CIFAR-10 dataset, it is verified on the VGG classification model. The algorithm is implemented using the Pytorch deep learning framework. The model training runs on an Ubuntu server with two NVIDIA GTX 2080Ti GPUs. The learning rate of gradient descent SGD is 0.01, the momentum is set to 0.9, and the batchsize is set to 128.

[0085] The large model adopts the standard VGG-16. The structure of the compressed small model is obtained by cropping the channels of each convolutional layer of the standard VGG-16. That is, the number of channels of the 13 convolutional layers of the standard VGG-16 is [64, 64, 128, 128, 256, 256, 256, 512, 512, 512, 512, 512, 512].

[0086] The hyperparameters are set as: α=0.95 in formula 2, and t...

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 customized deep neural network model compression method and system based on cloud-edge cooperation, and the method comprises the steps of carrying out the classification of concerned classification and unconcerned classification of a network model based on a high-precision network model, trained for a long time, of a cloud end and the personalized demands of a user; and then, in combination with the concerned classification, the data set of the user and the model compression ratio, through a neural network model compression method based on knowledge arrangement, lightening the model, so that the scene requirement that the user edge node resources are limited is met, the model reasoning speed is increased, and meanwhile, certain model accuracy is ensured. Accordingto the model training method based on user attention classification requirements and the knowledge distillation technology, verification is carried out based on a public data set in a picture classification scene.

Description

technical field [0001] The invention belongs to the field of edge computing and deep learning, and in particular relates to a customized deep neural network model compression method and system based on cloud-edge collaboration. Background technique [0002] Deep learning is an important branch of artificial intelligence. Deep learning, through deep neural networks, is having an increasingly important influence in fields such as machine vision, NLP, and speech recognition. With the development of edge computing, deep learning begins to play a more important role in a wider range of hardware devices - that is, "edge empowerment", which enables many hardware in the Internet of Things to have more powerful identification, detection, and intelligent processing. ability. However, many hardware devices in edge computing scenarios have relatively weak computing power, and the inference speed of deep neural networks is relatively slow. How to optimize and accelerate deep neural net...

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): G06F9/50G06N3/04G06N3/063G06N3/08
CPCG06F9/5072G06N3/063G06N3/08G06N3/045
Inventor 梁松涛高丰杨涛施佩琦汪明军郁善金王晓江郑欢欢
Owner ZHEJIANG LAB
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
Eureka Blog
Learn More
PatSnap group products