PRVNet: A Novel Partially-Regularized Variational Autoencoders for Massive MIMO CSI Feedback

被引:16
作者
Hussien, Mostafa [1 ,2 ]
Kim Khoa Nguyen [1 ]
Cheriet, Mohamed [1 ]
机构
[1] Univ Quebec, Ecole Technol Super ETS, Quebec City, PQ, Canada
[2] Assiut Univ, Informat Technol Dept, Assiut, Egypt
来源
2022 IEEE WIRELESS COMMUNICATIONS AND NETWORKING CONFERENCE (WCNC) | 2022年
关键词
MIMO-OFDM; CSI Feedback; Autoencoders;
D O I
10.1109/WCNC51071.2022.9771642
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In a multiple-input multiple-output frequency-division duplexing (MIMO-FDD) system, the user equipment (UE) sends the downlink channel state information (CSI) to the base station to report link status. Due to the complexity of MIMO systems, the overhead incurred in sending this information negatively affects the system bandwidth. Although this problem has been widely considered in the literature, prior work generally assumes an ideal feedback channel. In this paper, we introduce PRVNet, a neural network architecture inspired by variational autoencoders (VAE) to compress the CSI matrix before sending it back to the base station under noisy channel conditions. Moreover, we propose a customized loss function that best suits the special characteristics of the problem being addressed. We also introduce an additional regularization hyperparameter for the learning objective, which is crucial for achieving competitive performance. In addition, we provide an efficient way to tune this hyperparameter using KL-annealing. Experimental results show the proposed model outperforms the benchmark models including two deep learning-based models in a noise-free feedback channel assumption. In addition, the proposed model achieves an outstanding performance under different noise levels for additive white Gaussian noise feedback channels.
引用
收藏
页码:2286 / 2291
页数:6
相关论文
共 19 条
[1]  
Chengbo L.I., 2009, TV Minimization by Augment. Lagrangian Altern. Dir. Algorithms
[2]   Downlink Training Techniques for FDD Massive MIMO Systems: Open-Loop and Closed-Loop Training With Memory [J].
Choi, Junil ;
Love, David J. ;
Bidigare, Patrick .
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING, 2014, 8 (05) :802-814
[3]   An iterative thresholding algorithm for linear inverse problems with a sparsity constraint [J].
Daubechies, I ;
Defrise, M ;
De Mol, C .
COMMUNICATIONS ON PURE AND APPLIED MATHEMATICS, 2004, 57 (11) :1413-1457
[4]  
Guo JJ, 2020, IEEE T WIREL COMMUN, V19, P2827, DOI [10.1109/TWC.2020.2968430, 10.1109/TNSE.2020.2997359]
[5]  
Hussien M, 2021, INT CONF NETW SER, P427, DOI 10.23919/CNSM52442.2021.9615563
[6]   Towards More Reliable Deep Learning-Based Link Adaptation for WiFi 6 [J].
Hussien, Mostafa ;
Ahmed, Mohammed F. A. ;
Dahman, Ghassan ;
Kim Khoa Nguyen ;
Cheriet, Mohamed ;
Poitau, Gwenael .
IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS (ICC 2021), 2021,
[7]   Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification [J].
He, Kaiming ;
Zhang, Xiangyu ;
Ren, Shaoqing ;
Sun, Jian .
2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 2015, :1026-1034
[8]  
Kingma D. P., 2013, Auto-encoding variational bayes
[9]  
Kingma DP, 2014, ADV NEUR IN, V27
[10]   Variational Autoencoders for Collaborative Filtering [J].
Liang, Dawen ;
Krishnan, Rahul G. ;
Hoffman, Matthew D. ;
Jebara, Tony .
WEB CONFERENCE 2018: PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE (WWW2018), 2018, :689-698