Optimization method for transverse federated learning

A technology of horizontal federation and optimization method, applied in neural learning methods, biological neural network models, etc., can solve problems such as failure to obtain expected results, complicated source data for horizontal federated learning participants, and difficulty in ensuring independent and identical distribution. The effect of protecting data privacy

Pending Publication Date: 2021-04-30
HANGZHOU DIANZI UNIV
View PDF0 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, the source data of the horizontal federated learning participants is complex, and it is difficult to guarantee the condition of independence and same distribution. However, the existing horizontal federated learning optimization method is based on the assumption of independent and identical distribution of data, and often cannot obtain the expected results.

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
  • Optimization method for transverse federated learning
  • Optimization method for transverse federated learning
  • Optimization method for transverse federated learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0048]This embodiment provides an optimization method for lateral federal learning, including the following steps:

[0049]S1: The acquisition characteristics is similar but different source data, simulating the realistic scene Industrial state, but also the data status of the different participants of the customer, and normalize the data;

[0050]S2: Data isomerism based on the non-independent single distribution of each client (participant) local data based on the SMOTE-NON-IID model policy is derived by the data isomer of non-Idependently and Identically Distributed, Non-Idependently and Identically Distributed, Non-IID;

[0051]S2-1: Classify the processed data according to the tag, classify the client according to the data tag, so that the number of clients is equal to the number of data, while ensuring that each client is available and only one label;

[0052]S2-2: Each client uses the SMOTE algorithm to generate synthetic data, and passes these data to the server (coordinator), and final...

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 belongs to the technical field of federated learning and big data, and particularly relates to an optimization method for transverse federated learning, which comprises the following steps: S1, enabling a client to acquire data and preprocess the data; s2, processing the data of the client through an SMOTE-Non-IID model; s3, performing SMOTE-Non-IID model training through a client server architecture in transverse federated learning in combination with a data generation environment; s4, obtaining an optimization model of the SMOTE-Non-IID, and obtaining an evaluation result through prediction; according to the invention, each client synthesizes the data by adopting the SMOTE algorithm and receives the synthesized data from other clients at the same time, so that the problem of data heterogeneity in federated learning is effectively solved, and relatively high prediction accuracy can be ensured.

Description

Technical field[0001]The invention belongs to the federal learning and large data technology, which relates to an optimization method for horizontal federal learning.Background technique[0002]In the process of artificial intelligence applications, companies and other organizations are difficult to get a lot of quality data for the AI ​​model effect. Different organizations often do not provide their respective data for aggregation, resulting in data in the form of island; and at the same time, domestic and foreign regulatory environments have gradually strengthened data management and introduce relevant privacy protection policies. Therefore, the integration of data between different organizations will be very challenging, how to solve data fragmentation and data isolation in compliance with more strict privacy protection regulations is the current artificial intelligence researcher and practitioners The primary challenge.[0003]In order to solve the currently existing data island an...

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): G06N3/08
CPCG06N3/084
Inventor 张忠良程慧慧陈琼雒兴刚蔡灵莎
Owner HANGZHOU DIANZI 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