Federated Learning Client Training Lifecycle With Termination Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
After federated learning training is completed, existing systems leave clients in a waiting state, occupying space and computing power without a defined processing mechanism.
Innovation Solution
Implement a mechanism where a device sends a message indicating termination or suspension of federated learning training, allowing the client to perform operations such as stopping local training, updating or deleting the model, and saving resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the client stays in a waiting state for the next round of federated learning training, then the client can be ready for subsequent training rounds, but space and computing power are occupied without a defined processing mechanism
Solution Approach 1:
The patent extracts the client from the waiting state by introducing a termination message mechanism. When the server sends a termination message indicating completion of federated learning training, the client exits the waiting state and performs cleanup operations, thereby releasing occupied computing power and space resources while maintaining the ability to participate in future training rounds.
2Reliability
If the client stays in a waiting state for the next round of federated learning training, then the client can be ready for subsequent training rounds, but space is occupied without a defined processing mechanism
Solution Approach 1:
The patent extracts the client from the waiting state by introducing a termination message mechanism. When the server sends a termination message indicating completion of federated learning training, the client exits the waiting state and performs cleanup operations, thereby releasing occupied computing power and space resources while maintaining the ability to participate in future training rounds.
3Device complexity
If no processing mechanism is defined after federated learning training ends, then the system structure remains simple, but resources are wasted due to clients staying in waiting state
Solution Approach 1:
The patent implements a feedback mechanism where the server sends a termination message to clients after federated learning training ends. This feedback loop enables clients to understand when training is complete and triggers appropriate cleanup actions, efficiently releasing computing power and space resources without significantly increasing system structural complexity.
Data Source
AI summary
Embodiments of this application disclose a model training method, a terminal, and a network-side device. The model training method in the embodiments of this application includes: receiving, by a first device, a first message from a second device, where the first message is used to indicate termination or suspension of federated learning training; and performing, by the first device, a first operation based on the first message, where the first device includes a federated learning client, and the second device includes a federated learning server.