Go to AAAI 2025 Workshop AI4WCN homepageHow Stochastic Geometry and Machine Learning Coexist in Wireless Networks: Collaboration or Competition?
Keywords: Stochastic geometry, wireless communications
Abstract: Stochastic Geometry (SG) and Machine Learning (ML) are considered two highly effective methods for designing and evaluating the performance of next-generation large-scale wireless networks. SG is a model-driven approach that leverages previous research experience and mathematical derivations, while ML is a data-driven approach that learns from available datasets. Recently, it is indeed surprising that these two distinct methods have been frequently interacting in the field of wireless communication, coexisting through cooperation and competition. In existing studies, three types of interactions have been identified: (1) ML methods optimize the SG-based point process (PP) to bring it closer to the expected distribution; (2) The SG framework can be utilized to analyze or accelerate the convergence of ML-based methods; (3) ML can substitute SG in symmetric scenarios for performance evaluation and has been proven to be an evolution of SG for real-world scenarios. Furthermore, we design a novel and comprehensive case study called terrain-based coverage estimation, which encompasses all three types of interactions.
Submission Number: 1
Loading