Machine Learning Based Performance Approximation for Millimeter-Wave Cellular Networks

Millimeter-wave multicell networks require a complex mathematical analysis or a long simulation time due to such intrinsic properties as blockage, harsh propagation loss, and beamforming. In this regard, this paper proposes the approximation method of signal-to-interference-plus-noise-ratio (SINR) distributions using polynomial logistic functions and artificial neural networks, and it demonstrates that the proposed method provides a quick yet considerably accurate performance via computer simulations.