A wireless resource scheduling method and system based on knowledge distillation

By using knowledge distillation technology in wireless resource scheduling, the local sensitivity knowledge of the teacher's neural network model is transferred to the student's network model, which solves the problem of lightweight networks deviating from the optimal solution when there are small disturbances in the channel environment, and realizes efficient and accurate wireless resource scheduling.

CN122395733APending Publication Date: 2026-07-14JINGCHU UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
JINGCHU UNIV OF TECH
Filing Date
2026-04-13
Publication Date
2026-07-14

Smart Images

  • Figure CN122395733A_ABST
    Figure CN122395733A_ABST
Patent Text Reader

Abstract

The application provides a wireless resource scheduling method and system based on knowledge distillation, and the method comprises the following steps: obtaining training samples, which comprise network state data, optimal scheduling decision labels and gradient response data; training a teacher neural network model based on the training samples, the teacher neural network model being configured to output a first scheduling decision based on the network state data and to obtain a teacher gradient response output based on the gradient response data; training a student neural network model with a smaller number of parameters than the teacher neural network model through knowledge distillation; during the knowledge distillation training, constructing a distillation loss term according to the teacher gradient response output and a student gradient response output of the student neural network model; deploying the trained student neural network model to a wireless network edge node; and outputting real-time scheduling decisions of wireless resources based on the student neural network model. The application improves the timeliness and accuracy of wireless resource scheduling.
Need to check novelty before this filing date? Find Prior Art