Deep Learning in Network-Level Performance Prediction Using Cross-Layer Information

Abstract

Wireless communication networks are conventionally designed in model-based approaches through utilizing performance metrics such as spectral efficiency and bit error rate. However, from the perspectives of wireless service operators, network-level performance metrics such as the 5%-tile user data rate and network capacity are far more important. Unfortunately, it is difficult to mathematically compute such network-level performance metrics in a model-based approach. To cope with this challenge, this work proposes a data-driven machine learning approach to predict these network-level performance metrics by utilizing customized deep neural networks (DNN). More specifically, the proposed approach capitalizes on cross-layer information from both the physical (PHY) layer and the medium access control (MAC) layer to train customized DNNs, which was considered impossible for the conventional model-based approach. Furthermore, a robust training algorithm called weighted co-teaching (WCT) is devised to overcome the noise existing in the network data due to the stochastic nature of the wireless networks. Extensive simulation results show that the proposed approach can accurately predict two network-level performance metrics, namely user average throughput (UAT) and acknowledgment (ACK)/negative acknowledgment (NACK) feedback with great accuracy.