Federal learning method based on network unloading

A learning method and network technology, applied in neural learning methods, ensemble learning, biological neural network models, etc., can solve the problem of transmission rate, delay, reliability, uneven computing power of communication terminals, and large computing load of terminals. and other problems to achieve the effect of reducing communication overhead, alleviating transmission and computing pressure, and reducing communication pressure.

Pending Publication Date: 2021-08-10
浙江凡双科技有限公司
View PDF0 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] However, there are still many challenges in using federated learning in wireless communication scenarios
First of all, in the traditional federated learning application scenario, the data between the server and the terminal usually uses a limited connection, so the communication overhead is negligible. High requirements, the communication overhead introduced by federated learning will have an impact on network performance, affecting the original transmission rate, delay, reliability and other performance
Secondly, federated learning will generate a large amount of computing overhead on the communication terminal, especially in the scenario of the Internet of Everything, the computing capabilities of various communication terminals are uneven, and local computing will bring a large computing load to the terminal

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
  • Federal learning method based on network unloading
  • Federal learning method based on network unloading
  • Federal learning method based on network unloading

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0045] This embodiment provides a federated learning method based on network offload, which is applied to a wireless communication network with multiple intelligent terminals and an edge server, and the terminals and the edge server jointly train an artificial intelligence network model.

[0046] In this embodiment, the artificial intelligence network model is divided into two parts, which are trained on the mobile terminal side and the edge server side respectively. Select some mobile terminals for each round of training. In particular, a mobile terminal in a good state is selected according to the network environment where the mobile terminal is located, the transmission load and the calculation load.

[0047] In this embodiment, in order to select a mobile terminal in a good current state for training, after each round of training, a new mobile terminal in a good state will be reselected. Specifically, the process of the federated learning method based on network offloadin...

Embodiment 2

[0059] This embodiment provides a federated learning method based on network offload, which is applied to a wireless communication network with multiple mobile terminals and an edge server, and the mobile terminal and the edge server jointly train an artificial intelligence network model.

[0060] In this embodiment, the artificial intelligence network model is divided into two parts, which are trained on the mobile terminal side and the edge server side respectively. In particular, the edge server selects the number of network layers to be offloaded from the terminal to the edge server for training according to the computing capabilities of the mobile terminals participating in the training in the network.

[0061] Specifically, a federated learning method based on network offloading selects the number of network layers that are offloaded from the terminal to the edge server for training. The process is as follows image 3 shown, including:

[0062] 301. The edge server spec...

Embodiment 3

[0070] This embodiment provides a federated learning method based on network offload, which is applied to a wireless communication system with multiple edge servers, and uses data in terminals under all edge servers to jointly train an artificial intelligence network. like Figure 4 As shown in the figure, there is a cloud server, several edge servers and a large number of terminals in the system. The terminals and edge servers jointly train an artificial intelligence network and complete the gradient averaging in the cloud server.

[0071] In this embodiment, the artificial intelligence network model is divided into two parts, which are trained on the terminal side and the edge server side respectively. Each edge server completes the gradient average of its lower terminal, and transmits the gradient to the cloud server, and the cloud server completes the gradient average of the entire network.

[0072] Specifically, a federated learning method based on network offload is app...

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 federated learning method based on network unloading, which is used for improving the performance of the federated learning method while ensuring (improving) the privacy of terminal data. The method comprises the following steps: in a wireless communication system with a plurality of terminals and an edge server, jointly training an artificial intelligence network by using the terminals and the edge server; compared with a traditional federal learning working mode, carrying out the training of partial networks by unloading a terminal to an edge server; in the training process, selecting a terminal which is good in wireless communication network environment and small in calculation and transmission load to participate in training of the network model. According to the invention, the calculation load of the communication terminal is effectively reduced, and the communication load caused by a federated learning method is reduced.

Description

technical field [0001] The invention relates to the field of artificial intelligence and communication, in particular to a federated learning method based on network offloading. Background technique [0002] In recent years, artificial intelligence (AI) technology has developed rapidly and has been widely used, and more complex and cutting-edge artificial intelligence technology has been expected to be applied in many fields, including driverless cars, medical care, finance, etc. Data is the foundation of artificial intelligence technology. However, in most industries, due to issues such as industry competition, privacy security, and complex administrative procedures, data often exists in the form of isolated islands. In reality, it is very difficult to integrate data scattered in various places and institutions. The federated learning technology can effectively deal with the above difficulties and use the data in the form of isolated islands for learning. [0003] The Fed...

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): G06F9/50G06N3/04G06N3/08G06N20/20
CPCG06F9/5072G06F2209/509G06N3/04G06N3/084G06N20/20
Inventor 吴哲奕张春林许重九邢焕赵琛迪
Owner 浙江凡双科技有限公司
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