핵심 개념
Dynamic spectrum access requires effective spectrum occupancy detection, improved by federated learning algorithms.
초록
I. Abstract
Dynamic spectrum access crucial for radio communication.
Effective spectrum occupancy detection key.
Machine learning enhances detection.
Federated learning used for distributed detection.
II. Introduction
Dynamic Spectrum Access (DSA) system needed.
Spectrum sensing crucial for occupancy detection.
Machine learning improves detection quality.
III. Data Collection
Measurements conducted for spectrum occupancy detection.
Data collected for simulation using federated learning.
IV. Simulation Setup
Data balanced for machine learning.
Data divided into subsets for federated learning simulation.
V. Simulation Results
Federated learning improves efficiency in spectrum occupancy detection.
Logistic regression and neural network tested.
VI. Conclusions
Federated learning shows potential for reliable detection.
Further research needed for environmental impact on models.
통계
"The average efficiency of the presented federated learning algorithm was 94.51%, whereas the average efficiency of the sensors (without federated learning) was 92.74%."
"The average efficiency of the presented federated learning algorithm was 96.46%, whereas the average efficiency of the sensors (excluding federated learning) was 95.63%."
"With two broken (in the same way) sensors, the average efficiency of the presented federated learning algorithm was 96.34%, whereas the average efficiency of the sensors (excluding federated learning) was 77.79%."
인용구
"Federated learning shows potential for using in spectrum occupancy detection systems."
"Using even such a simple algorithm as averaging model coefficients improves system reliability."