Deep Transfer Learning-Based Downlink Channel Prediction for FDD Massive MIMO Systems

Yang, Yuwen; Gao, Feifei; Zhong, Zhimeng; Ai, Bo; Alkhateeb, Ahmed

doi:10.1109/tcomm.2020.3019077

Cited by 144 publications

(88 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…10 shows the energy efficiency results for both full-duplex and half-duplex schemes. Moreover, we consider the utilization of transfer block architecture to enhance the energy efficiency results by reducing the number of RF chains 11 . As proven in Section V, the total achievable rate performance remains the same whether the transfer block architecture is utilized or not.…”

Section: Energy Efficiencymentioning

confidence: 99%

“…Nonetheless, the minimization problem for the current form of MSE j depends on the effective SI channel H SI,j , which is not available at the node j. According to the SI channel model given in (11), the SI channel has two parts indicating the LoS paths (i.e., the near-field SI channel) and NLoS paths (i.e., far-field SI channel). After applying the antenna separation based SIC, the NLoS paths become more dominant compared to the LoS paths (i.e., H SI,j = F r,j H SI,j F t,j ≈ F r,j H NLoS,j F t,j ).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Full-Duplex mmWave Massive MIMO Systems: A Joint Hybrid Precoding/Combining and Self-Interference Cancellation Design

Koç

Le‐Ngoc

2021

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

Millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) systems have been considered as one of the primary candidates for the fifth generation (5G) and beyond 5G wireless communication networks to satisfy the ever-increasing capacity demands. Full-duplex technology can further enhance the advantages of mmWave massive MIMO systems. However, the strong self-interference (SI) is the major limiting factor in the full-duplex technology. Hence, this paper proposes a novel angular-based joint hybrid precoding/combining (AB-JHPC) technique for the full-duplex mmWave massive-MIMO systems. Our primary goals are listed as: (i) improving the self-interference cancellation (SIC), (ii) increasing the intended signal power, (iii) decreasing the channel estimation overhead, (iv) designing the massive MIMO systems with a low number of RF chains. First, the RF-stage of AB-JHPC is developed via slow time-varying angle-of-departure (AoD) and angle-of-arrival (AoA) information. A joint transmit/receive RF beamformer design is proposed for covering (excluding) the AoD/AoA support of intended (SI) channel. Second, the BB-stage of AB-JHPC is constructed via the reduced-size effective intended channel. After using the well-known singular value decomposition (SVD) approach at the BB-stage, we also propose a new semi-blind minimum mean square error (S-MMSE) technique to further suppress the residual SI power by using AoD/AoA parameters. Thus, the instantaneous SI channel knowledge is not needed in the proposed AB-JHPC technique. Finally, we consider a transfer block architecture to minimize the number of RF chains. The numerical results demonstrate that the SI signal is remarkably canceled via the proposed AB-JHPC technique. It is shown that AB-JHPC achieves 85.7 dB SIC and the total amount of SIC almost linearly increases via antenna isolation techniques. We observe that the proposed full-duplex mmWave massive MIMO systems double the achievable rate capacity compared to its half-duplex counterpart as the antenna array size increases and the transmit/receive antenna isolation improves. Moreover, the proposed S-MMSE algorithm provides considerably high capacity than the conventional SVD approach.Index Terms-Full-duplex, massive MIMO, millimeter wave communications, hybrid precoding, hybrid combining, low CSI overhead, RF chain, semi-blind MMSE, energy efficiency, imperfect angle information.

show abstract

Section: Energy Efficiencymentioning

confidence: 99%

mentioning

confidence: 99%

Full-Duplex mmWave Massive MIMO Systems: A Joint Hybrid Precoding/Combining and Self-Interference Cancellation Design

Koç

Le‐Ngoc

2021

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

show abstract

“…We choose to introduce this layer before the deep learning layers as the neural network designs are driven by applications. [7] Sleep monitoring CNN+RNN Adversarial domain adaptation Zheng et al 2019 [5] Gesture recognition CNN+RNN HCF Fhager et al 2019 [8] Gesture recognition CNN Transfer learning Tamzeed et al 2019 [11] Gesture recognition CNN+LSTM Zero-shot learning Yang et al 2020 [12] Gesture recognition CNN+LSTM Teacher student network Wang et al 2019 [13] Pose estimation U-Net+attention Adversarial domain adaptation Guan et al 2020 [6] Imaging CNN GAN Yang et al 2019 [14] Channel prediction FNN Transfer learning…”

Section: Application Layermentioning

confidence: 99%

Enhancing RF Sensing with Deep Learning: A Layered Approach

et al. 2021

View full text Add to dashboard Cite

In recent years, (Radio Frequency) RF sensing has gained increasing popularity due to its pervasiveness, low-cost, non-intrusiveness, and privacy preservation. However, realizing the promises of RF sensing is highly non-trivial, given typical challenges such as multipath and interference. One potential solution leverages deep learning to build direct mappings from RF domain to target domains, hence avoiding complex RF physical modeling. While earlier solutions exploit only simple feature extraction and classification modules, an emerging trend adds functional layers on top of elementary modules for more powerful generalizability and flexible applicability. To better understand this potential, this paper takes a layered approach to summarize RF sensing enabled by deep learning. Essentially, we present a four-layer framework: physical, backbone, generalization, and application. While this layered framework provides readers a systematic methodology for designing deep interpreted RF sensing, it also facilitates us to make improvement proposals and to hint future research opportunities.

show abstract

“…Since the OSNR features are similar for higher-order QAM, direct-learning may cause the negative transfer between different modulation formats. To avoid the negative transfer among unrelated tasks, it is considered as a single-source domain transfer to a multi-source domain transfer problem rather than a direct-transfer problem [21] as shown in Fig. 1 (c).…”

Section: Principle Of the Cascaded Neural Networkmentioning

confidence: 99%

Joint Modulation Format Identification and OSNR Monitoring Using Cascaded Neural Network With Transfer Learning

Zhang

et al. 2021

IEEE Photonics J.

View full text Add to dashboard Cite

We propose a cascaded neural network (NN) to simultaneously identify the modulation formats and monitor the optical-signal-to-noise ratio (OSNR). In the second-level network, it is a single deep NN (DNN) rather than multiple sub-networks, which makes the architecture more compact and can save the resource for real implementation. However, since the data set is constituted from all modulation formats, the universality can be guaranteed but not for the accuracy and the complexity. To accelerate the estimation process and improve the accuracy, we introduce the transfer learning (TL) and reconstruct the data set with a part from all of the modulation formats for universality and another part from a specific modulation format for TL to pursue higher accuracy. In the experiment, we compare the proposed cascaded single neural network (CSNN) with or without TL, cascaded multiple neural networks (CMNN) and adaptive multi-task learning (MTL) for MFI and OSNR monitoring. In the first-level NN, all of the three schemes can achieve the accuracy of MFI as 100 %. In the second-level NN, the CSNN with TL (TL-CSNN) can significantly improve the training speed and decline the RMSE of 0.19 dB compared with CSNN without TL. The TL-CSNN also has faster convergence speed and is more stable compared with CMNN and adaptive MTL.

show abstract

Deep Transfer Learning-Based Downlink Channel Prediction for FDD Massive MIMO Systems

Cited by 144 publications

References 30 publications

Full-Duplex mmWave Massive MIMO Systems: A Joint Hybrid Precoding/Combining and Self-Interference Cancellation Design

Full-Duplex mmWave Massive MIMO Systems: A Joint Hybrid Precoding/Combining and Self-Interference Cancellation Design

Enhancing RF Sensing with Deep Learning: A Layered Approach

Joint Modulation Format Identification and OSNR Monitoring Using Cascaded Neural Network With Transfer Learning

Contact Info

Product

Resources

About