Federated Learning Incentive Calculation for Privacy-Preserving Model Updates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In federated learning, clients participating in the model training do not receive adequate incentives based on their contribution to the model's accuracy, leading to a lack of motivation for significant contributions while maintaining data privacy and security.

Innovation Solution

An information processing system that calculates and provides incentives to clients based on their contribution degree to the global model by receiving local models or output data from multiple nodes, updating the global model, and evaluating the contribution of each local model or output data to the updated global model.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If federated learning is used to train global model with data privacy and security, then data privacy and security are improved, but clients do not receive adequate incentives based on their contribution to model accuracy

Engineering Contradiction:
Improvedata privacy and securityVSAvoidclient motivation and incentive
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The system calculates and provides feedback to each client about their contribution degree to the global model accuracy. This feedback mechanism motivates clients by showing them their specific impact, while the federated learning architecture maintains data privacy and security by keeping data local. The feedback loop closes by using this contribution information to adjust incentives or recognition.

Inventive Principle:
Principle #23Feedback

2Reliability

If local data is kept distributed to ensure data privacy and security, then data privacy and security are improved, but contribution degree calculation becomes more difficult

Engineering Contradiction:
Improvedata privacy and securityVSAvoidcontribution degree calculation
Core Design Contradiction:
ReliabilityVSDifficulty of detecting and measuring

Solution Approach 1:

The system introduces an intermediary mechanism (contribution degree calculation module) that mediates between the distributed local data and the global model. This intermediary calculates contribution degrees by analyzing model updates, gradients, or performance metrics without requiring access to actual local data, thus maintaining privacy while enabling accurate contribution measurement.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system replaces direct mechanical data aggregation with a computational approach using model parameters, gradients, or performance indicators to infer contribution degrees. Instead of mechanically summing up data contributions, the system uses mathematical operations on model representations to substitute for direct data access, enabling contribution measurement without data centralization.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If clients participate in federated learning without adequate incentive, then data privacy and security are maintained, but client participation and contribution are insufficient

Engineering Contradiction:
Improvedata privacy and securityVSAvoidclient participation and contribution
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system implements a feedback mechanism that calculates and communicates contribution degrees to participating clients. This feedback serves as an incentive by recognizing and rewarding clients based on their actual contribution to model accuracy, thereby increasing motivation for participation while the federated learning framework maintains data privacy and security.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system changes the parameter of client motivation by introducing contribution degree as a measurable and rewardable parameter. By transforming the abstract concept of contribution into a quantifiable parameter that can be tracked and rewarded, the system increases client participation and contribution while maintaining the privacy-preserving federated learning architecture.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4379610A1Information processing apparatus, node, information processing method, carrier means, and information processing system
Publication Date: 2024.06.05 RICOH CO LTD
  • EP4379610A1 patent drawingFigure 1
  • EP4379610A1 patent drawingFigure 2
  • EP4379610A1 patent drawingFigure 3

AI summary

An information processing apparatus, a node, an information processing method, a carrier means, and an information processing system. The information processing apparatus receives information indicating a local model or output data, from each of a plurality of nodes, the information indicating the local model being obtained by learning a local data processed by the node based on a global model, the output data being obtained by inputting shared data to the local model, updates the global model based on a plurality of the information indicating the local model or a plurality of the output data received from the plurality of nodes, and calculates a contribution degree of each of the plurality of local models or each of the plurality of output data to the updated global model.