Deep Learning Based MIMO Communications

O’Shea, Timothy J.; Erpek, Tugba; Clancy, T. Charles

doi:10.48550/arxiv.1707.07980

Cited by 45 publications

(85 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also plot the curves for the SVD with bit allocation as a baseline. This is similar to [4], [5] that we all have parallel sub-channels. The difference is we choose the best modulation schemes among BPSK and M -quadrature amplitude modulation (M -QAM)to reach the minimum BER over all of the sub-channels while the total bits transmitted is N s .…”

Section: A Performance Of Svd In Daesupporting

confidence: 63%

“…• We test the performance of per-bit input and per-bit output regression and compare it with the commonlyused one-hot input and one-hot output [4], [6]. From our simulation, bit input is a good candidate for DAE especially when the number of input bits is 2 and 4.…”

Section: A Contributionsmentioning

confidence: 99%

“…massive MIMO [13], [14] and so on. In general, DL is a suitable learning and optimization tool and has a high potential in improving the computational complexity, which can improve energy efficiency when such systems are deployed [4]. This is because DL-based schemes can benefit from the parallel hardware on graphical processing units (GPUs) and advanced processors such as accelerated processing unit (APU), tensor processing unit (TPU), neural network processing unit (NPU), and so on.…”

Section: B Related Workmentioning

confidence: 99%

“…• One-hot input: In this case, the bit vector is initially converted to a one-hot vector v o and then transmitted to the DAE. The one-hot input has already been used in existing DAE designs like [4], [6]. More specifically, the kth one-hot vector in the codebook, v…”

Section: ) Transmitted Informationmentioning

confidence: 99%

“…DAE design is more challenging when it comes to multipleinput multiple-output (MIMO) systems. DAE in MIMO communication systems was first investigated in [4] where a MIMO spatial multiplexing system using DAE is designed with the perfect knowledge of channel state information at the transmitter (CSIT). The DAE takes advantage of CSIT to cancel interference at the receiver, and can be seen as decorrelator/interference nuller the same role singular-value decomposition (SVD) plays in MIMO systems [5].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

SVD-Embedded Deep Autoencoder for MIMO Communications

Zhang¹,

Vaezi²,

O’Shea³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Using a deep autoencoder (DAE) for end-to-end communication in multiple-input multiple-output (MIMO) systems is a novel concept with significant potential. DAE-aided MIMO has been shown to outperform singular-value decomposition (SVD)-based precoded MIMO in terms of bit error rate (BER). This paper proposes embedding left-and right-singular vectors of the channel matrix into DAE encoder and decoder to further improve the performance of MIMO spatial multiplexing. SVD-embedded DAE largely outperforms theoretic linear precoding in terms of BER. This is remarkable since it demonstrates that the proposed DAEs have significant potential to exceed the limits of current system design by treating the communication system as a single, end-to-end optimization block. Based on the simulation results, at SNR=10dB, the proposed SVD-embedded design can achieve BER nearly 10 −5 and reduce the BER at least 10 times compared with existing DAE without SVD, and up to 18 times improvement compared with theoretical linear precoding. We attribute this to the fact that the proposed DAE can match the input and output as an adaptive modulation structure with finite alphabet input. We also observe that adding residual connections to the DAE further improves the performance.• We propose and design an SVD-embedded DAE network.Precoding and pre-processing at Tx and Rx have been added as one layer with non-trainable weights. This design can achieve BER nearly 10 −5 at SNR=10dB, and can reduce the BER 10 to 30 times compared with the

show abstract

Section: A Performance Of Svd In Daesupporting

confidence: 63%

Section: A Contributionsmentioning

confidence: 99%

Section: B Related Workmentioning

confidence: 99%

Section: ) Transmitted Informationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

SVD-Embedded Deep Autoencoder for MIMO Communications

Zhang¹,

Vaezi²,

O’Shea³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Machine learning assisted adaptive LDPC coded system design and analysis

Xie,

El‐Hajjar,

2023

IET Communications

View full text Add to dashboard Cite

This paper proposes a novel machine learning (ML) assisted low‐latency low density parity check (LDPC) coded adaptive modulation (AM) system, where short block‐length LDPC codes are used. Conventional adaptive modulation and coding (AMC) system includes fixed look‐up table method, which is also called inner loop link adaptation (ILLA) and outer loop link adaptation (OLLA). For ILLA, the adaptive capability is achieved by switching the modulation and coding modes based on a look‐up table using signal‐to‐noise ratio (SNR) thresholds at the target bit error rate (BER), while OLLA builds upon the ILLA method by dynamically adjusting the SNR thresholds to further optimize the system performance. Although both improve the system overall throughput by switching between different transmission modes, there is still a gap to optimal performance as the BER is comparatively far away from the target BER. Machine learning (ML) is a promising solution in solving various classification problems. In this work, the supervised learning based k‐nearest neighbours (KNN) algorithm is invoked for choosing the optimum transmission mode based on the training data and the instantaneous SNR. This work focuses on the low‐latency communications scenarios, where short block‐length LDPC codes are utilized. On the other hand, given the short block‐length constraint, we propose to artificially generate the training data to train our ML assisted AMC scheme. The simulation results show that the proposed ML‐LDPC‐AMC scheme can achieve a higher throughput than the ILLA system while maintaining the target BER. Compared with OLLA, the proposed scheme can maintain the target BER while the OLLA fails to maintain the target BER when the block length is short. In addition, when considering the channel estimation errors, the performance of the proposed ML‐LDPC‐AMC maintains the target BER, while the ILLA system's BER performance can be higher than the target BER.

show abstract

Trainable Projected Gradient Detector for Massive Overloaded MIMO Channels: Data-Driven Tuning Approach

et al. 2019

View full text Add to dashboard Cite

This paper presents a deep learning-aided iterative detection algorithm for massive overloaded multiple-input multiple-output (MIMO) systems where the number of transmit antennas n is larger than that of receive antennas m. Since the proposed algorithm is based on the projected gradient descent method with trainable parameters, it is named the trainable projected gradient-detector (TPG-detector). The trainable internal parameters, such as the step-size parameter, can be optimized with standard deep learning techniques, i.e., the back propagation and stochastic gradient descent algorithms. This approach is referred to as data-driven tuning, and ensures fast convergence during parameter estimation in the proposed scheme. The TPG-detector mainly consists of matrix-vector product operations whose computational cost is proportional to mn for each iteration. In addition, the number of trainable parameters in the TPG-detector is independent of the number of antennas. These features of the TPG-detector result in a fast and stable training process and reasonable scalability for large systems. Numerical simulations show that the proposed detector achieves a comparable detection performance to those of existing algorithms for massive overloaded MIMO channels, e.g., the state-of-the-art IW-SOAV detector, with a lower computation cost.INDEX TERMS massive MIMO, overloaded MIMO, detection algorithm, deep learning VOLUME 4, 2016

show abstract

Deep Learning Based MIMO Communications

Cited by 45 publications

References 14 publications

SVD-Embedded Deep Autoencoder for MIMO Communications

SVD-Embedded Deep Autoencoder for MIMO Communications

Machine learning assisted adaptive LDPC coded system design and analysis

Trainable Projected Gradient Detector for Massive Overloaded MIMO Channels: Data-Driven Tuning Approach

Contact Info

Product

Resources

About