Intelligent MIMO Detection With Momentum-Induced Unfolded Layers

Sangbu Yun,Seungsik Moon,Yo-Seb Jeon,Youngjoo Lee

DOI: https://doi.org/10.1109/lwc.2023.3348933

IF: 6.3

2024-03-09

IEEE Wireless Communications Letters

Abstract:In this letter, we present a novel deep-learning-based network for MIMO symbol detection, referred to as a momentum-induced detection network, MomentNet. Inspired by the projected gradient descent algorithm, our proposed architecture integrates an additional momentum component to the previous deep-learning detectors. By eliminating redundant values, we have successfully reduced the number of training parameters by half from the baseline network, compensating for the increased computational burden introduced by the incorporation of momentum parameters. Simulation results reveal that the proposed network achieves near-Maximum Likelihood symbol error rate performance with low computational complexity.

computer science, information systems,telecommunications,engineering, electrical & electronic