Federated Model Update Using Difference Consistency Feedback

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In federated learning, the fixed or ascending/descending training information settings can lead to suboptimal model updates, affecting the accuracy of training as the optimal upload time for client models is not dynamically adjusted based on previous training results.

Innovation Solution

A method for calculating first training information using difference consistency information from multiple clients to optimize model updates, including formulas for determining training batch quantity and learning rate adjustments based on consistency metrics and performance weights.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If fixed or ascending/descending training information is used, then the service device can simply deliver training parameters, but the client may miss optimal upload occasions and training accuracy deteriorates

Engineering Contradiction:

Improvesimplicity of training information deliveryVSAvoidtraining accuracy

Core Design Contradiction:

Ease of operationVSManufacturing precision

Solution Approach 1:

The patent applies dynamics by transitioning from fixed or predetermined training information to dynamically adjusted training parameters. The service device calculates training information based on real-time difference consistency information from multiple clients, allowing the training parameters (learning rate, batch size) to adapt dynamically to the current state of model updates across the federated network, thereby resolving the contradiction between simplicity and accuracy.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements feedback by using difference consistency information calculated from multiple clients' model differences to adjust training parameters. The service device receives difference information from clients, calculates consistency metrics, and uses this feedback to dynamically determine optimal training information for the next round, creating a closed-loop system that improves training accuracy while maintaining operational simplicity.

Inventive Principle:

Principle #23Feedback

2Manufacturing precision

If training information is dynamically adjusted based on difference consistency, then training accuracy improves, but the service device complexity increases

Engineering Contradiction:

Improvetraining accuracyVSAvoidservice device complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies self-service by enabling the service device to automatically calculate difference consistency information and determine training parameters based on the uploaded difference information from clients. The system performs self-adjustment without external intervention, where the service device autonomously processes consistency calculations and generates appropriate training information, improving accuracy while the automation masks the underlying computational complexity.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent uses parameter changes by adjusting training parameters (learning rate, batch size) based on calculated difference consistency information. The service device modifies these parameters dynamically according to the consistency level detected from multiple clients' model differences, allowing the system to adapt to varying training conditions and improve accuracy through parameter optimization.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If training information is set according to ascending or descending rules, then the service device can follow a simple pattern, but the client misses optimal upload occasions related to previous training results

Engineering Contradiction:

Improveadaptability to training resultsVSAvoidcomplexity of training information calculation

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the training information into separate components (learning rate, batch size) that can be independently adjusted based on difference consistency information. This allows the system to segment the complexity of parameter optimization and adjust each parameter appropriately based on the current training state, improving adaptability while managing complexity through modular adjustment.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12572850B2Method for implementing model update and device thereof

Publication Date: 2026.03.10 HUAWEI TECH CO LTD

AI summary

A method for implementing model update is provided and used in a federated update framework. The method includes: a service device receives difference information uploaded by at least two clients. The service device performs calculation based on the difference information uploaded by the at least two clients, to obtain first difference consistency information, where the first difference consistency information indicates a consistency degree of the difference information uploaded by the at least two clients. The service device performs calculation based on the first difference consistency information, to obtain first training information, where the first training information is used to train a third model, and the third model is obtained by updating the first model by the service device based on the difference information uploaded by the at least two clients. The service device sends the first training information to the at least two clients.