Federal learning data participant contribution value calculation and excitation method based on block chain

A technology of blockchain and contribution value, applied in computing, computing models, machine learning, etc., can solve problems such as unfairness, data privacy leakage of learning models, inaccurate evaluation methods and reward mechanisms, etc.

Active Publication Date: 2021-05-11
ZHEJIANG UNIV +1
View PDF12 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Second, data privacy and data security issues
Evaluation methods and reward mechanisms are inaccurate and unfair
At the same time, this method will include the model information in the federated learning process, the trained model of the participants, the test data, and the contribution, etc., in the blockchain. Due to the openness of the blockchain, there may be learning models and even data privacy. risk of leakage

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 data participant contribution value calculation and excitation method based on block chain
  • Federal learning data participant contribution value calculation and excitation method based on block chain
  • Federal learning data participant contribution value calculation and excitation method based on block chain

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0057] Such as figure 1 As shown, a blockchain-based federated learning data participant contribution value calculation and incentive method, the method includes steps:

[0058] First, the federated learning model distributes the model to each data participant, and the data participant receives the distributed model, uses the local data based on its own local data to train the distributed model, and obtains the local training model;

[0059] Perform feature extraction on the local training model of the data participant, calculate the gradient change of the local training model, and obtain model update data;

[0060] Such as figure 2 As shown, the gradient change adopts the gradient descent algorithm (Gradient Descent Optimization), according to the formula of gradient descent In the case of simplified learning rate effects, the gradient change of participant i's local model is distributed by the federated model of the round using participant i's local data d i The gradien...

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 federal learning data participant contribution value calculation and excitation method based on a block chain, and the method comprises the steps: carrying out the feature extraction of a local training model of a data participant, and obtaining model updating data; enabling a data demander to decode the model update data to obtain an intermediate verification model; enabling the data demand side to evaluate the intermediate verification model in batches to obtain evaluation indexes; calculating the contribution degree of the data participant according to the evaluation index; distributing excitation values to the data participants according to the contribution degrees of the data participants; and forming a reward queue by the excitation values and the addresses of the data participants, and putting the reward queue on the block chain. According to the method, the contribution of the data demander to the data participant is evaluated and calculated, and then the contribution of each party is used as a distribution basis for reward distribution, so that the fairness is ensured, and meanwhile, the enthusiasm of the data owner to participate in federated learning is improved; and the data owners are effectively encouraged and attracted to participate in federal learning.

Description

technical field [0001] The invention relates to the technical field of federated learning, in particular to a blockchain-based federated learning data participant contribution value calculation and incentive method. Background technique [0002] In recent years, artificial intelligence technology has developed rapidly and has been widely used in various industries. Artificial intelligence driven by the big data environment has entered a golden period of development. Some potential problems currently exist may cause big data-driven artificial intelligence technology to enter a trough in development, which needs to be faced and resolved: First, the problem of data sources, including limited data volume and data quality problems. In many industries, data exists in the form of data islands, and there are many obstacles to the integration of data between industries and even within industries. The second is data privacy and data security issues. How to effectively integrate and ...

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): G06N20/00G06F16/27G06Q20/06
CPCG06N20/00G06F16/27G06Q20/065
Inventor 孙凌云王鑫余芸萧展辉杨漾邵明陈波敖知琪周泽宝任昊文崔焱甘杉邹文景孙刚杨晓雪
Owner ZHEJIANG UNIV
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