Related Experiment Video
Updated: Sep 16, 2025
13:44
Simulation, Fabrication and Characterization of THz Metamaterial Absorbers
Published on: December 27, 2012
15.5K
Interpretable artificial intelligence for modulated metasurface antenna design using SHAP and MLP
Amrollah Amini1, Ali Moshiri2, Mohammad Amin Chaychi Zadeh1
1School of Advanced Technologies, Iran University of Science and Technology, Tehran, 16846-13114, Iran.
Scientific Reports
|July 5, 2025
Summary
This study introduces an interpretable AI framework using SHAP and MLP to predict antenna radiation metrics. The method enhances metasurface antenna design by identifying key features for improved sidelobe level and beamwidth prediction.
Area of Science:
- Electromagnetics and Applied Physics
- Artificial Intelligence and Machine Learning
Background:
- Metasurface-based leaky-wave antennas are crucial for modern telecommunications, imaging, and radar due to their compact and efficient design.
- Precise control over antenna radiation metrics like sidelobe level (SLL) and half-power beamwidth (HPBW) is essential for system performance.
Purpose of the Study:
- To develop an interpretable artificial intelligence (AI) framework for predicting key radiation metrics of metasurface antennas.
- To leverage explainable AI (XAI) techniques for feature engineering and model optimization.
Main Methods:
- Integration of SHapley Additive exPlanations (SHAP) with a multi-layer perceptron (MLP) for feature importance analysis.
- Development of a multi-task neural network for joint prediction of SLL and HPBW.
- Feature engineering based on SHAP insights to improve predictive accuracy.
Main Results:
- SHAP analysis identified dominant features and interactions influencing antenna radiation.
- The multi-task neural network achieved near-perfect accuracy in SLL prediction.
- Significant improvements were observed in HPBW estimation accuracy.
Conclusions:
- Interpretable AI enhances understanding of antenna behavior and guides model refinement.
- The proposed framework offers a generalizable approach for optimizing metasurface antenna systems.
- Explainable AI integration leads to superior predictive performance in antenna design.
