18159571. Massively Scalable, Resilient, and Adaptive Federated Learning System simplified abstract (Huawei Technologies Co., Ltd.)
Massively Scalable, Resilient, and Adaptive Federated Learning System
Organization Name
Inventor(s)
Masood Seyed Mortazavi of Plano TX (US)
Massively Scalable, Resilient, and Adaptive Federated Learning System - A simplified explanation of the abstract
This abstract first appeared for US patent application 18159571 titled 'Massively Scalable, Resilient, and Adaptive Federated Learning System
Simplified Explanation
The abstract describes a federated learning system that allows multiple clients to contribute model updates to a central repository. The system includes scalable queues to receive these model update contributions, a model repository to store the model, and a configuration repository to store model policies.
- The system allows multiple clients to contribute updated model parameters to improve the model.
- The model repository stores the model and receives updates based on the updated model parameters.
- The configuration repository stores model policies, including an update threshold that determines when to initiate a model update.
- Hierarchical aggregators update the model based on the updated model parameters from the clients and the update threshold.
Potential Applications
- Collaborative machine learning: Allows multiple clients to contribute to a shared model without sharing their data.
- Privacy-preserving learning: Protects sensitive data by keeping it on the client side and only sharing model updates.
- Distributed learning: Enables training of models across a large number of clients, improving scalability and efficiency.
Problems Solved
- Data privacy: Clients can contribute to the model without sharing their raw data, preserving privacy.
- Scalability: The system can handle a large number of clients and efficiently update the model.
- Centralized model management: The model repository and configuration repository provide a centralized way to store and manage the model and its updates.
Benefits
- Improved model accuracy: By aggregating model updates from multiple clients, the system can improve the overall model accuracy.
- Privacy protection: Clients' data remains on their devices, reducing privacy concerns.
- Scalability and efficiency: The system can handle a large number of clients and efficiently update the model, making it suitable for distributed learning scenarios.
Original Abstract Submitted
A federated learning system is disclosed. The system includes scalable queues configured to receive model update contributions from a plurality of clients. The model update contributions contain updated model parameters. The system also includes a model repository configured to store a model for access by a plurality of clients and receive the model with updates based on the updated model parameters. The system also includes a configuration repository configured to store model polices including an update threshold indicating how many responses need to be received from the plurality of clients to initiate an update of the model. The system also includes hierarchical aggregators configured to update the model based on the updated model parameters from the plurality of clients and based on the update threshold.
