Server Apparatus for Federated Learning with Multiplex Branches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In federated learning, the risk of information leakage is high due to the sharing of local models among clients, making it difficult to obtain a model optimized for each client while maintaining privacy.
Innovation Solution
A server apparatus and client apparatus configuration that receives and transmits local model parameters and weights for neural network models with multiplex branches, allowing each client to learn a global model parameter based on its own data without sharing local models, thereby reducing information leakage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If local models are shared among all clients to achieve personalized federated learning, then each client can obtain a model optimized for its own data, but the risk of information leakage increases
Solution Approach 1:
The patent extracts only the necessary model parameters (local model parameters and branch weights) from the complete local models and transmits them to the server. This extraction approach allows clients to receive personalized model updates without sharing their entire local models, thereby achieving model personalization while minimizing information leakage risk.
Solution Approach 2:
The patent enables each client to maintain its own local model with personalized characteristics by receiving customized model updates from the server. Each client's local model has unique quality characteristics adapted to its specific data distribution, while the server aggregates information from all clients through their transmitted parameters rather than their complete models.
2Measurement precision
If complete local models are transmitted to the server for aggregation, then the global model can be updated with full client information, but the amount of data transmission and privacy risk increase
Solution Approach 1:
The patent extracts only the essential model parameters (local model parameters and branch weights) from the complete local models for transmission to the server. This selective extraction reduces data transmission volume significantly while preserving the necessary information for accurate global model aggregation.
Solution Approach 2:
The patent segments the model information into two distinct components: local model parameters and branch weights. This segmentation allows the system to transmit only the necessary portions of model information to the server, reducing overall data transmission requirements while maintaining aggregation accuracy.
3Adaptability or versatility
If personalized federated learning is implemented through local model sharing, then each client achieves optimized performance, but the system complexity and communication overhead increase
Solution Approach 1:
The patent extracts and transmits only the necessary model parameters and weights rather than complete local models. This extraction approach simplifies the communication protocol and reduces the complexity of model sharing operations while still enabling each client to achieve personalized model optimization through the received updates.
Data Source
AI summary
A server apparatus includes: a receiving unit configured to receive, from each of a plurality of client apparatuses performing federated learning of a neural network model having multiplex branches capable of performing different operations on a common input, a local model parameter of each of the multiplex branches and a weight for each branch used in superposing outputs from the respective multiplex branches; a calculating unit configured to calculate a parameter of a global model based on the local model parameter and the weight received by the receiving unit; and a transmitting unit configured to transmit the parameter calculated by the calculating unit to the client apparatuses.