Federal model fragmentation inference method and device based on dynamic bandwidth perception
By employing a dynamic bandwidth-aware federated model fragmentation inference method, and utilizing a lightweight bandwidth prediction model and multidimensional state vectors to optimize model fragment size and acquisition strategy, this approach solves the transmission blocking and latency problems caused by network bandwidth fluctuations in federated learning, achieving efficient model fragment transmission and task execution.
Patent Information
- Application Number
- CN202610873126.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-17
- Publication Date
- 2026-07-17
AI Technical Summary
Existing federated learning methods are prone to blocking and delays when the network bandwidth fluctuates significantly, resulting in low overall inference efficiency and an inability to adapt to changes in network conditions.
By employing a dynamic bandwidth-aware federated model fragmentation inference method, a lightweight bandwidth prediction model is used to predict network state, optimize model fragment size, and dynamically adjust fragment acquisition strategy based on multidimensional state vectors and fragmentation decision model, thereby determining source nodes and transmitting model fragments in the model fragmentation interaction network.
It effectively solves the problem of low transmission efficiency caused by the inability of fixed strategies to adapt to dynamic environmental changes, greatly reduces invalid transmission and waiting time, and ensures the timeliness of model fragment transmission and the continuity of task execution.
Smart Images
Figure CN122420315A_ABST