Designing Interpretable AI-Driven Propagation Models for Advanced Wireless Networks
Basit A. Zaidi, Shuja Ansari · 2024
This study draws inspiration from successful applications in biomedical data analysis, where guiding neural networks (NNs) with domain knowledge has proven effective in enhancing model performance through the integration of expert-derived representations [3]. While existing models in this domain excel in interpretability by aligning learned features with established domain expertise, our proposed method takes a unique approach. We introduce a novel domain knowledge-guided deep neural network that goes beyond the conventional application of domain knowledge. This pioneering effort seamlessly integrates domain knowledge into neural networks specifically designed for radio propagation modeling. The approach aims to enhance interpretability in this context, presenting a promising avenue for advancing the understanding and performance of neural networks in wireless communication scenarios.