Unlock instant, AI-driven research and patent intelligence for your innovation.

Model updating method, terminal and server

A model update and server technology, applied in the field of artificial intelligence, can solve the problems of large user knowledge forgetting and low accuracy of aggregation models, and achieve the effect of improving performance

Pending Publication Date: 2022-03-25
HUAWEI TECH CO LTD +1
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For scenarios with large differences in user data, there is a problem that the aggregated model forgets user knowledge and the accuracy of the aggregated model is low.

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
  • Model updating method, terminal and server
  • Model updating method, terminal and server
  • Model updating method, terminal and server

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0091] The embodiment of the present application provides a method for updating models, which is used to implement model aggregation, so that the aggregated model retains the user's knowledge to a greater extent, so as to improve the performance of the aggregated model. In addition, a model aggregation strategy is also proposed, and the group aggregation mode is adopted for multiple models, which further improves the performance of the aggregated model.

[0092] For ease of understanding, the following briefly introduces some technical terms involved in the embodiments of this application:

[0093] 1. Federated learning: a distributed machine learning algorithm that does not involve the exchange of private local data of each device while training on multiple terminals or edge devices. Local data is the data generated by the user during the use of the terminal, which may include private and sensitive data. Compared with the traditional centralized machine learning method, whic...

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 model updating method which is used for a federal learning scene, and provides a new weighting coefficient calculation method for user data distribution. According to the method provided by the embodiment of the invention, the terminal sends the trained model and the distribution characteristics of the local data to the cloud, and the distribution characteristics of the local data are used for replacing a uniform weighting coefficient to realize model aggregation, so that the aggregated model keeps the knowledge of a user to a greater extent, and the performance of the aggregated model is improved. In addition, a model aggregation strategy is provided, a grouping aggregation mode is adopted for multiple models, and the performance of the aggregated models is further improved.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, in particular to a method for updating a model, a terminal and a server. Background technique [0002] Federated learning is a distributed machine learning algorithm. While training on multiple terminals or edge devices, it does not involve the exchange of private data between each device, which can protect user privacy. Federated learning adopts the architecture of device-cloud collaboration, which consists of centralized cloud nodes and distributed terminal devices. One iteration of device-cloud collaboration includes device-side training, model upload, cloud aggregation, and model distribution. After the end-side model is trained, it is uploaded to the cloud for aggregation, processing, updating, and other operations, and then merged into a new model and distributed to the end-side device. [0003] In the existing device-cloud collaboration technology, when models a...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/214
Inventor 苏上超王勇博李斌薛向阳陈院林黄一宁
Owner HUAWEI TECH CO LTD