Federated Learning Edge Machine Collaboration System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial edge machines generate vast amounts of data but lack effective tools for secure and credible data sharing and application, leading to 'data islands' that hinder optimization and collaboration, with centralized cloud computing exacerbating communication delays and security concerns.
Innovation Solution
A system for collaboration and optimization of edge machines based on federated learning, which includes model parameter assignment and training units, scenario feature optimization, and encryption methods like homomorphic encryption to securely share and aggregate data, promoting intelligent machine collaboration and data value conversion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If centralized cloud computing is used for data collection and model training, then model training capability is improved, but data privacy security deteriorates and communication delays increase
Solution Approach 1:
The patent segments the centralized cloud computing system into distributed federated learning systems at edge machines. Each edge machine independently trains local models using its own data, then shares only model parameters (not raw data) with the global system. This segmentation enables model training capability while preserving data privacy security by keeping data localized.
Solution Approach 2:
The patent introduces an intermediary mechanism where only encrypted model parameters are transmitted between edge machines and the global system, rather than raw data. This intermediary approach allows model training capability while preventing data leakage, as the intermediary (encrypted parameters) cannot be used to reconstruct original data.
2Measurement precision
If all edge machine data is collected for centralized training, then model accuracy is improved, but communication overhead and delays worsen
Solution Approach 1:
The patent extracts the data transmission step from the model training process. Instead of transmitting raw data from all edge machines to the cloud for training, only essential model parameters are extracted and shared. This reduction in transmitted data volume significantly decreases communication overhead and delays while maintaining model accuracy through federated aggregation.
Solution Approach 2:
The patent applies partial action by having each edge machine train local models independently using its own data first, then contribute only the necessary model parameters to the global system. This partial approach (local training + parameter sharing) achieves model accuracy without requiring complete data collection, thereby reducing communication time.
3Adaptability or versatility
If data is shared among edge machines for collaboration, then system intelligence is improved, but data security risks worsen
Solution Approach 1:
The patent uses copying by transmitting encrypted copies of model parameters instead of original data. Each edge machine creates encrypted copies of its learned model parameters and shares them with the global system. This copying mechanism enables system intelligence through knowledge sharing while maintaining data security, as the copied parameters cannot be reverse-engineered to obtain original data.
Solution Approach 2:
The patent applies parameter changes by transforming raw data into model parameters through the training process, and then encrypting these parameters before transmission. This parameter transformation and encryption approach enables collaborative learning (improving system intelligence) while protecting data security, as the parameters exist in a different form and are protected by encryption.
Data Source
AI summary
A system for collaboration and optimization of edge machines based on federated learning is provided. The system includes R federated learning systems, R≥1, a model parameter assignment unit, and model training and optimizing units. The model parameter assignment unit is configured to assign initial parameters for federated learning to the Mi edge machines, receive intermediate model parameters, and aggregate and update the received intermediate model parameters to obtain new model parameters. The model training and optimizing units are configured to train, on the basis of the initial parameters and respective operating data, local operating models, transmit the intermediate model parameters obtained after training to the model parameter assignment unit, and obtain a system collaborative operating model according to the new model parameters.