Meta Learning-Based MIMO Detectors: Design, Simulation, and Experimental Test

Zhang, Jing; He, Yunhua; Li, Yuwen; Wen, Chao-Kai; Jin, Shi

doi:10.1109/twc.2020.3030882

Cited by 46 publications

(25 citation statements)

References 38 publications

(103 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…EPNet [19] and Meta-ViterbiNet [20] adopt a meta-learning framework. EPNet uses the meta-learned LSTM [21] optimizer introduced in [22].…”

Section: Related Workmentioning

confidence: 99%

MetaSSD: Meta-Learned Self-Supervised Detection

Park¹,

Ok²,

Jeon³

et al. 2022

Preprint

View full text Add to dashboard Cite

Deep learning-based symbol detector gains increasing attention due to the simple algorithm design than the traditional model-based algorithms such as Viterbi and BCJR. The supervised learning framework is often employed to predict the input symbols, where training symbols are used to train the model. There are two major limitations in the supervised approaches: a) a model needs to be retrained from scratch when new train symbols come to adapt to a new channel status, and b) the length of the training symbols needs to be longer than a certain threshold to make the model generalize well on unseen symbols. To overcome these challenges, we propose a meta-learning-based self-supervised symbol detector named MetaSSD. Our contribution is two-fold: a) meta-learning helps the model adapt to a new channel environment based on experience with various meta-training environments, and b) self-supervised learning helps the model to use relatively less supervision than the previously suggested learning-based detectors. In experiments, MetaSSD outperforms OFDM-MMSE with noisy channel information and shows comparable results with BCJR. Further ablation studies show the necessity of each component in our framework.

show abstract

“…EPNet [19] and Meta-ViterbiNet [20] adopt a meta-learning framework. EPNet uses the meta-learned LSTM [21] optimizer introduced in [22].…”

Section: Related Workmentioning

confidence: 99%

MetaSSD: Meta-Learned Self-Supervised Detection

Park¹,

Ok²,

Jeon³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…whereX denotes the decision on the RV X. Rearranging (12) and canceling f Y (y), we obtain the likelihood ratio…”

Section: A Learning-based Framework For the Proposed Mimo Soft Detection Schemementioning

confidence: 99%

“…On the other hand, MMNEt in [11] is an adaptive neural-network based detection scheme tailored to realistic channels with spatial correlation and is suitable for online training, where the training is performed for each coherence time instead of offline training where the module could be trained in advance for a large number of coherence times by generating random channel coefficients. In the same context of online training and motivated by practical implementation, EPNet [12] was proposed to perform signal detection by unfolding the expectation propagation (EP) algorithm and training the damping factors. To support coded systems, a neural-network MIMO detector with impairments was proposed in [13], where the detection algorithm design is based upon projected gradient descent iterations for MIMO-OFDM systems.…”

Section: Introductionmentioning

confidence: 99%

Data-Driven Precoded MIMO Detection Robust to Channel Estimation Errors

Mayouche

Martins

Chatzinotas

et al. 2021

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

We study the problem of symbol detection in downlink coded multiple-input multiple-output (MIMO) systems with precoding and without the explicit knowledge of the channel-state information (CSI) at the receiver. In this context, we investigate the impact of imperfect CSI at the transmitter (CSIT) on the detection performance. We first model the CSIT degradation based on channel estimation errors to investigate its impact on the detection performance at the receiver. To mitigate the effect of CSIT deterioration at the latter, we propose learningbased techniques for hard and soft detection that use downlink precoded pilot symbols as training data. We note that these pilots are originally intended for signal-to-interference-plus-noise ratio (SINR) estimation. We validate the approach by proposing a lightweight implementation that is suitable for online training using several state-of-the-art classifiers. We compare the biterror rate (BER) and the runtime complexity of the proposed approaches where we achieve superior detection performance in harsh channel conditions while maintaining low computational requirements. Specifically, numerical results show that severe CSIT degradation impedes the correct detection when a conventional detector is used. However, the proposed learning-based detectors can achieve good detection performance even under severe CSIT deterioration, and can yield 4-8 dB power gain for BER values lower than 10 −4 when compared to the classic linear minimum mean square error (MMSE) detector.

show abstract

“…Meta learning is an emerging learning technique which uses previous knowledge and experience to guide the learning process of new tasks [12]. It aims at designing a network that can, through learning, quickly adapt to the new environment through a relatively small amount of new training data.…”

Section: Introductionmentioning

confidence: 99%

Intelligent MIMO Detection Using Meta Learning

Huo

et al. 2022

IEEE Wireless Commun. Lett.

View full text Add to dashboard Cite

In a K-best detector for multiple-input-multiple-output (MIMO) systems, the value of K needs to be sufficiently large to achieve near-maximum-likelihood (ML) performance. By treating K as a variable that can be adjusted according to a fitting function of some learnable coefficients, an intelligent MIMO detection network based on deep neural networks (DNN) is proposed to reduce complexity of the detection algorithm with little performance degradation. In particular, the proposed intelligent detection algorithm uses meta learning to learn the coefficients of the fitting function for K to circumvent the problem of learning K directly. The idea of network fusion is used to combine the learning results of the meta learning component networks. Simulation results show that the proposed scheme achieves near-ML detection performance while its complexity is close to that of linear detectors. Besides, it also exhibits strong ability of fast training.

show abstract

Meta Learning-Based MIMO Detectors: Design, Simulation, and Experimental Test

Cited by 46 publications

References 38 publications

MetaSSD: Meta-Learned Self-Supervised Detection

MetaSSD: Meta-Learned Self-Supervised Detection

Data-Driven Precoded MIMO Detection Robust to Channel Estimation Errors

Intelligent MIMO Detection Using Meta Learning

Contact Info

Product

Resources

About