EVM Loss: A Loss Function for Training Neural Networks in Communication Systems

Abstract

Neural networks and their application in communication systems are receiving growing attention from both academia and industry. The authors note that there is a disconnect between the typical objective functions of these neural networks with regards to the context in which the neural network will eventually be deployed and evaluated. To this end, a new loss function is proposed and shown to increase the performance of neural networks when implemented in a communication system compared to previous methods. It is further shown that a `split complex’ approach used by many implementations can be improved via formalisation of the `concatenated complex’ approach described herein. Experimental results using the orthogonal frequency division multiplexing (OFDM) and spectrally efficient frequency division multiplexing (SEFDM) modulation formats with varying bandwidth compression factors over a wireless visible light communication (VLC) link validate the efficacy of the proposed method in a real system, achieving the lowest error vector magnitude (EVM), and thus bit error rate (BER), across all experiments, with a 5 dB to 10 dB improvement in the received symbols EVM overall compared to the baseline implementation, with bandwidth compressions down to 40% compared to OFDM, resulting in a spectral efficiency gain of 67%.

Machine Learning for Physical Layer in 5G and beyond Wireless Networks: A Survey

“…Various techniques enhance the physical layer performance of 5G; for example, ref. [220] proposed a loss function to increase the performance of neural networks in a communication system, and [221] proposed a multi-antenna and multi-subcarrier channel state information (CSI)-based novel channel sounder architecture to achieve an accuracy better than 75 cm for line of sight (LoS) for indoor user positioning in three dimensions. In [222], the author proposed an ANN-based novel Adaptive Modulation and Coding (AMC) scheme to estimate the signal-to-noise power ratio (SNR) to determine the optimal MCS with a low calculation complexity, and [223] proposed ML-based peak-to-average power ratio (PAPR) reduction using the optimal hyperparameter function and efficient approximation for the downlink channel of mMIMO with an OFDM signal.…”

Editorial to the Special Issue on “Visible Light Communications, Networking, and Sensing”

“…[7], in the paper "Utilization of an OLED-Based VLC System in Office, Corridor, and Semi-Open Corridor Environments", focus on the unique organic LED panel transmitters placed on walls in various indoor environments, based on the 3D ray tracing simulations that the new statistics of rootmean-square delay the spread, from which optical path losses are derived. The next paper by S. Stainton et al [8] on "A Loss Function for Training Neural Networks in Communication Systems" proposes a novel error vector magnitude (EVM) loss method, which seeks to reconnect the disparity between the evaluation performed when training a neural network and the evaluation of the overall communication system. The results experimentally validated the capabilities of the method, which consistently obtained the lowest EVM when deployed in a real experimental setup, using orthogonal frequency division multiplexing (OFDM) and spectrally efficient frequency division multiplexing with a varying bandwidth.…”