Channel Modeling Based on Transformer Symbolic Regression for Inter-Satellite Terahertz Communication

Channel modeling is crucial for inter-satellite terahertz communication system design. The conventional method involves manually constructing a mathematical channel model, which is labor-intensive, and using a neural network directly as a channel model lacks interpretability. This paper introduces a...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Applied sciences 2024-04, Vol.14 (7), p.2929
Hauptverfasser: He, Yuanzhi, Sheng, Biao, Li, Zhiqiang
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Channel modeling is crucial for inter-satellite terahertz communication system design. The conventional method involves manually constructing a mathematical channel model, which is labor-intensive, and using a neural network directly as a channel model lacks interpretability. This paper introduces a channel modeling approach based on symbolic regression. It is the first time that using transformer neural networks as the implementation tool of symbolic regression to generate the mathematical channel model from the channel data directly. It can save manpower and avoid the interpretability issue of using neural networks as a channel model. The feasibility of the proposed method is verified by generating a free space path loss model from simulation data in the terahertz frequency band.
ISSN:2076-3417
2076-3417
DOI:10.3390/app14072929