Researchers have developed an explainable machine learning (XML) framework designed to improve the transparency and reliability of path loss modeling for millimeter-wave (mmWave) and sub-terahertz (sub-THz) wireless communication systems. Conventional machine learning models frequently function as black-box systems, which presents challenges for interpretability and physical reliability in deployment-critical infrastructure. This study addresses these limitations by integrating physics-guided deep neural network (DNN) regression with SHapley Additive exPlanations (SHAP), providing a more trustworthy approach for modeling complex signal propagation behaviors that exceed the capabilities of traditional empirical models.
The proposed framework systematically evaluates path loss by incorporating physical consistency verification and explanation stability analysis. By employing these techniques, the researchers aim to ensure that the learned relationships between propagation parameters align with established wireless communication principles. This methodology allows the system to not only predict propagation behavior across diverse environments but also to quantify the stability of these explanations, thereby enhancing the deployment-oriented utility of the channel modeling.
To validate the framework, the team utilized publicly available multi-band propagation datasets covering frequency ranges from 28 GHz to 140 GHz. The analysis examined the influence of several key propagation variables, including the separation distance between transmitter and receiver, carrier frequency, antenna heights, and line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. By analyzing these factors, the researchers sought to isolate the specific drivers of signal attenuation in high-frequency environments.
Experimental results indicate that propagation distance and carrier frequency consistently function as the dominant drivers of path loss across both mmWave and sub-THz regimes. Conversely, factors such as antenna heights and LOS/NLOS conditions were found to exert environment-dependent corrective effects on the models. The study demonstrated that the XML framework successfully maintains physically meaningful monotonic attenuation behavior with respect to both distance and frequency across the vast majority of evaluated samples.
In comparative performance tests, the XML framework was evaluated against conventional empirical models, including the close-in and alpha–beta–gamma models, as well as baseline machine learning approaches such as support vector regression, random forest, and standard black-box DNNs. The researchers report that their framework achieves competitive prediction accuracy while simultaneously providing substantially improved interpretability and physical transparency compared to existing black-box methods.
The findings offer actionable insights that could support network planning, digital-twin-assisted wireless optimization, and the data-driven calibration of analytical channel models. By prioritizing reproducibility and deployment reliability, this research provides a pathway for integrating machine learning into the design of future 6G communication systems, moving toward models that are both performant and physically explainable.
Subject of Research: Explainable machine learning framework for mmWave and sub-THz wireless path loss modeling.
Article Title: Explainable machine learning for path loss modeling in mmWave and Sub-THz communication systems
Article References: Mula, S. S. R., & Pamula, V. K. (2026). Explainable machine learning for path loss modeling in mmWave and Sub-THz communication systems. Journal on Wireless Communications and Networking. https://doi.org/10.1186/s13638-026-02668-9
Image Credits: AI Generated
DOI: 10.1186/s13638-026-02668-9
Keywords: Machine Learning, mmWave, Sub-THz, Path Loss Modeling, Explainable AI, Wireless Communication, 6G, Physics-Guided Neural Networks, Signal Propagation, Digital Twins, Channel Modeling, SHAP, Network Planning, Propagation Datasets, Deployment Reliability
Cite Scienmag News
Teresa Odom. (November 18, 2026). Explainable Machine Learning Enhances Path Loss Modeling for Future Wireless Systems. Scienmag. https://scienmag.com/explainable-machine-learning-enhances-path-loss-modeling-for-future-wireless-systems/
Teresa Odom. "Explainable Machine Learning Enhances Path Loss Modeling for Future Wireless Systems." Scienmag, 18 November 2026, https://scienmag.com/explainable-machine-learning-enhances-path-loss-modeling-for-future-wireless-systems/. Accessed 18 November 2026.
Teresa Odom. "Explainable Machine Learning Enhances Path Loss Modeling for Future Wireless Systems." Scienmag. November 18, 2026. https://scienmag.com/explainable-machine-learning-enhances-path-loss-modeling-for-future-wireless-systems/








