Biao Sheng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Biao Sheng has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Yuanzhi He
Sun Yat-sen University
1 shared publication - Zhiqiang Li
Institute of Systems Engineering, Academy of Military Sciences, Beijing, China
1 shared publication
Latest Publications
- Open Access
- 30 March 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (7) , 2929
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 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.
- 30 March 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (7) , 2929
Open AccessChannel 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 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.