Yuwei Zeng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yuwei Zeng has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Chuan Zhang
LEADS, the National Mobile Communications Research Laboratory, and the Frontiers Science Center for Mobile Information Communication and Security, Southeast University, Nanjing, China
8 shared publications - Yongming Huang
National Mobile Communications Research Laboratory, Southeast University, Nanjing, China
6 shared publications - Xiaohu You
Southeast University
4 shared publications - Yingmeng Ge
LEADS, the National Mobile Communications Research Laboratory, and the Frontiers Science Center for Mobile Information Communication and Security, Southeast University, Nanjing, China
4 shared publications - You You
LEADS, the National Mobile Communications Research Laboratory, and the Frontiers Science Center for Mobile Information Communication and Security, Southeast University, Nanjing, China
4 shared publications - Zaichen Zhang
National Mobile Communications Research Laboratory, Frontiers Science Center for Mobile Information Communication and Security, Southeast University, Nanjing, China
4 shared publications
Latest Publications
- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 1489-1493
ATLBO-Q: Adaptive TLBO-Based Quantization for MMSE MIMO Detection
Quantization is of significance for multiple-input multiple-output (MIMO) detector design in wireless communication systems. Existing uniform quantization methods suffer from high bitwidth redundancy and inefficient hardware resource utilization, while machine-learning-based automatic quantization frameworks introduce considerable computational burden. In this letter, an adaptive teaching-learning-based-optimization (TLBO)-based quantization (ATLBO-Q) framework is proposed for minimum mean square error MIMO detectors, where Cholesky decomposition is utilized to eliminate floating-point matrix inversion operations. The bitwidth assignments are optimized iteratively through an adaptive teacher-learner phase, avoiding the complex training procedures. Simulation results demonstrate that the proposed ATLBO-Q achieves significant reduction in average bitwidth and computational complexity while maintaining near-floating-point bit error rate performance, thereby attaining a desirable trade-off between the detection precision and hardware efficiency.
- 19 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 1489-1493
ATLBO-Q: Adaptive TLBO-Based Quantization for MMSE MIMO Detection
Quantization is of significance for multiple-input multiple-output (MIMO) detector design in wireless communication systems. Existing uniform quantization methods suffer from high bitwidth redundancy and inefficient hardware resource utilization, while machine-learning-based automatic quantization frameworks introduce considerable computational burden. In this letter, an adaptive teaching-learning-based-optimization (TLBO)-based quantization (ATLBO-Q) framework is proposed for minimum mean square error MIMO detectors, where Cholesky decomposition is utilized to eliminate floating-point matrix inversion operations. The bitwidth assignments are optimized iteratively through an adaptive teacher-learner phase, avoiding the complex training procedures. Simulation results demonstrate that the proposed ATLBO-Q achieves significant reduction in average bitwidth and computational complexity while maintaining near-floating-point bit error rate performance, thereby attaining a desirable trade-off between the detection precision and hardware efficiency.
- 2 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 750-754
DeepSAQ: Deep Learning-Driven Sensitivity-Aware Quantization for MMSE MIMO Detection
Hardware-efficient quantization strategies are imperative for massive MIMO detectors. However, conventional methods such as unified quantization and manual tuning respectively suffer from bitwidth redundancy and prohibitive exploration overhead. To address these challenges, this letter proposes a deep learning-driven sensitivity-aware quantization (DeepSAQ) framework, which includes: 1) A gradient-driven framework with a Bayesian optimized loss function to minimize fractional bitwidths; 2) Sensitivity-aware tuning to compress the fractional bitwidth redundancy; 3) Range-extended integral bitwidths to further improve the quantization efficiency. Applying to the unfolded MMSE-wNSA detection, the proposed framework delivers floating-point comparable performance but reduces 50.04% average bitwidths and 50.47% computational complexity compared to the unified quantization. Benchmarking with the recent art, the proposed framework succeeds in squeezing up to 18.32% bitwidths and 21.23% computational complexity in 8 $\times $ 128 64-QAM MIMO systems.
- 2 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 750-754
DeepSAQ: Deep Learning-Driven Sensitivity-Aware Quantization for MMSE MIMO Detection
Hardware-efficient quantization strategies are imperative for massive MIMO detectors. However, conventional methods such as unified quantization and manual tuning respectively suffer from bitwidth redundancy and prohibitive exploration overhead. To address these challenges, this letter proposes a deep learning-driven sensitivity-aware quantization (DeepSAQ) framework, which includes: 1) A gradient-driven framework with a Bayesian optimized loss function to minimize fractional bitwidths; 2) Sensitivity-aware tuning to compress the fractional bitwidth redundancy; 3) Range-extended integral bitwidths to further improve the quantization efficiency. Applying to the unfolded MMSE-wNSA detection, the proposed framework delivers floating-point comparable performance but reduces 50.04% average bitwidths and 50.47% computational complexity compared to the unified quantization. Benchmarking with the recent art, the proposed framework succeeds in squeezing up to 18.32% bitwidths and 21.23% computational complexity in 8 $\times $ 128 64-QAM MIMO systems.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 745-749
MD-GWO: A Meta-Heuristic-Driven Quantizer for Low-Complexity MIMO Detection
Effective quantization can reduce implementation redundancy in multiple-input multiple-output (MIMO) systems by achieving reduced bit-width while preserving system performance. However, the unified quantification (UQ) and current hybrid-precision quantization methods fail to achieve optimal bit-width reduction due to quantization coarseness or deficiencies in global search capabilities. To address these issues, this letter proposes a meta-heuristic-driven automatic quantization method for MIMO detection, named multi-population discrete grey wolf optimizer (MD-GWO). The proposed method introduces three key techniques: 1) a cosine-annealed convergence factor adaptation scheme, 2) a constrained fitness function that balances bit-width and performance loss, and 3) a multi-population cooperative optimization framework. Numerical results demonstrate the superior performance of MD-GWO when applied in the minimum mean square error (MMSE) and channel hardening-exploiting message passing (CHEMP) detectors under 16-QAM modulation. The method achieves an average fractional bit-width reduction of up to $\mathbf {61.63}\%$ and $\mathbf {32.31\%}$ in comparison to UQ and the state-of-the-art, respectively, without sacrificing system bit error rate (BER) performance.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15 (21622337) , 745-749
MD-GWO: A Meta-Heuristic-Driven Quantizer for Low-Complexity MIMO Detection
Effective quantization can reduce implementation redundancy in multiple-input multiple-output (MIMO) systems by achieving reduced bit-width while preserving system performance. However, the unified quantification (UQ) and current hybrid-precision quantization methods fail to achieve optimal bit-width reduction due to quantization coarseness or deficiencies in global search capabilities. To address these issues, this letter proposes a meta-heuristic-driven automatic quantization method for MIMO detection, named multi-population discrete grey wolf optimizer (MD-GWO). The proposed method introduces three key techniques: 1) a cosine-annealed convergence factor adaptation scheme, 2) a constrained fitness function that balances bit-width and performance loss, and 3) a multi-population cooperative optimization framework. Numerical results demonstrate the superior performance of MD-GWO when applied in the minimum mean square error (MMSE) and channel hardening-exploiting message passing (CHEMP) detectors under 16-QAM modulation. The method achieves an average fractional bit-width reduction of up to $\mathbf {61.63}\%$ and $\mathbf {32.31\%}$ in comparison to UQ and the state-of-the-art, respectively, without sacrificing system bit error rate (BER) performance.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (3) , 596-600
GausiQ: Generalized Automatic Hybrid-Precision Quantization for MIMO Detection
Automatic quantization generates efficient hybrid precision quantization schemes without manual effort, offering a promising approach for developing hardware-friendly MIMO detectors. However, unified quantization (UQ) and existing automatic quantization methods typically optimize the integral part (IP) and fractional part (FP) bitwidths separately, resulting in considerable bitwidth redundancy. To address it, we propose a generalized automatic hybrid-precision quantization (GausiQ) algorithm, employing two key techniques: 1) joint optimization of the IP and FP bitwidths based on DRL, and 2) reduction of the inter-variable interference through terminal state. The proposed GausiQ has been evaluated in a 16×32 16-QAM MIMO system, utilizing MMSE detection as a case study. Numerical results demonstrate that the incorporation of terminal state can significantly accelerate the convergence rate. Compared to UQ and the state-of-the-art, GausiQ has reduced the average bitwidth by 29.67% and 12.33% without penalizing BER performance.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (3) , 596-600
GausiQ: Generalized Automatic Hybrid-Precision Quantization for MIMO Detection
Automatic quantization generates efficient hybrid precision quantization schemes without manual effort, offering a promising approach for developing hardware-friendly MIMO detectors. However, unified quantization (UQ) and existing automatic quantization methods typically optimize the integral part (IP) and fractional part (FP) bitwidths separately, resulting in considerable bitwidth redundancy. To address it, we propose a generalized automatic hybrid-precision quantization (GausiQ) algorithm, employing two key techniques: 1) joint optimization of the IP and FP bitwidths based on DRL, and 2) reduction of the inter-variable interference through terminal state. The proposed GausiQ has been evaluated in a 16×32 16-QAM MIMO system, utilizing MMSE detection as a case study. Numerical results demonstrate that the incorporation of terminal state can significantly accelerate the convergence rate. Compared to UQ and the state-of-the-art, GausiQ has reduced the average bitwidth by 29.67% and 12.33% without penalizing BER performance.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (3) , 576-580
BANQ: BayesOpt-Based Automatic Non-Uniform Quantization for SCL Polar Decoding
Successive cancellation list (SCL) polar decoding is well-known for its outstanding error correction performance. To achieve a better balance between performance and complexity in SCL decoders, non-uniform quantization (NUQ) is commonly employed. NUQ strategically adjusts the quantization steps to improve the decoder’s performance within specific bitwidth constraints. However, existing NUQ schemes rely on the designers’ expertise and lack an automated design methodology. To address this issue, this letter formulates the NUQ optimization problem and solves it using Bayesian-optimization (BayesOpt), resulting in an automated NUQ scheme termed BANQ. Utilizing BayesOpt to fine-tune the quantization steps, BANQ empowers the SCL decoder to deliver enhanced error correction performance with comparable complexity. For a (512,410) 5G polar code, compared to uniform quantization (UQ), BANQ cuts bitwidth by 20% and bit operations (BOPs) by 19.5%, achieving near floating-point accuracy. Furthermore, a case study verifies that BANQ can be effectively adapted for 5G LDPC min-sum decoding.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (3) , 576-580
BANQ: BayesOpt-Based Automatic Non-Uniform Quantization for SCL Polar Decoding
Successive cancellation list (SCL) polar decoding is well-known for its outstanding error correction performance. To achieve a better balance between performance and complexity in SCL decoders, non-uniform quantization (NUQ) is commonly employed. NUQ strategically adjusts the quantization steps to improve the decoder’s performance within specific bitwidth constraints. However, existing NUQ schemes rely on the designers’ expertise and lack an automated design methodology. To address this issue, this letter formulates the NUQ optimization problem and solves it using Bayesian-optimization (BayesOpt), resulting in an automated NUQ scheme termed BANQ. Utilizing BayesOpt to fine-tune the quantization steps, BANQ empowers the SCL decoder to deliver enhanced error correction performance with comparable complexity. For a (512,410) 5G polar code, compared to uniform quantization (UQ), BANQ cuts bitwidth by 20% and bit operations (BOPs) by 19.5%, achieving near floating-point accuracy. Furthermore, a case study verifies that BANQ can be effectively adapted for 5G LDPC min-sum decoding.
- 14 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 13 (12) , 3355-3359
DL-AQUA: Deep-Learning-Based Automatic Quantization for MMSE MIMO Detection
Directly affecting both error performance and complexity, quantization is critical for MMSE MIMO detection. However, naively pruning quantization levels is not advised as it typically causes performance degradation. In practice, unified quantization (UQ) across different parameters based on empirical judgments or numerical simulations is mostly employed, as it circumvents sophisticated design and implementation efforts. Though UQ leaves room for performance/complexity improvements, manually fine-tuned quantization necessitates substantial efforts and is usually prohibitive for practical usage. To automate the quantization process, this letter first proposes a Deep-Learning based Automatic QUAntization (DL-AQUA) framework for MMSE detection. Multi-step optimization techniques, involving 1) gradient based search, 2) adaptive penalty based tuning, and 3) probability density function aware calibration, help to attain an agile yet fine-tuned quantization configuration. For 8×128 MIMO detection, DL-AQUA successfully achieves 35.06% quantization reduction than UQ without penalizing performance. Even benchmarking against the recent AHPQ technique, the relative gains are 14.53%. The robustness of DL-AQUA has been demonstrated across different antenna and SNR combinations. It should be noted that the immediate application of DL-AQUA is not limited to MMSE detection, but also extends to other quantization-sensitive DSP algorithms.
- 14 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 13 (12) , 3355-3359
DL-AQUA: Deep-Learning-Based Automatic Quantization for MMSE MIMO Detection
Directly affecting both error performance and complexity, quantization is critical for MMSE MIMO detection. However, naively pruning quantization levels is not advised as it typically causes performance degradation. In practice, unified quantization (UQ) across different parameters based on empirical judgments or numerical simulations is mostly employed, as it circumvents sophisticated design and implementation efforts. Though UQ leaves room for performance/complexity improvements, manually fine-tuned quantization necessitates substantial efforts and is usually prohibitive for practical usage. To automate the quantization process, this letter first proposes a Deep-Learning based Automatic QUAntization (DL-AQUA) framework for MMSE detection. Multi-step optimization techniques, involving 1) gradient based search, 2) adaptive penalty based tuning, and 3) probability density function aware calibration, help to attain an agile yet fine-tuned quantization configuration. For 8×128 MIMO detection, DL-AQUA successfully achieves 35.06% quantization reduction than UQ without penalizing performance. Even benchmarking against the recent AHPQ technique, the relative gains are 14.53%. The robustness of DL-AQUA has been demonstrated across different antenna and SNR combinations. It should be noted that the immediate application of DL-AQUA is not limited to MMSE detection, but also extends to other quantization-sensitive DSP algorithms.
- 15 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (7) , 10767-10771
A Deep-Learning-Aided Message Passing Detector for MIMO SC-FDMA
In this paper, a multiple-input multiple-output (MIMO) single carrier frequency division multiple access (SC-FDMA) detector is proposed, named simplified generalized approximate message passing (Sim-GAMP), which is derived from the existing low-complexity GAMP (LC-GAMP). To improve its BER performance, Sim-GAMP is further developed by means of deep learning (DL) techniques, which is called DL-GAMP in this work. Specifically, the layers of DL-GAMP are constructed by unfolding iterations of Sim-GAMP and introducing learnable parameters, thus contributing to better BER performance. Numerical results demonstrate that DL-GAMP can mitigate the performance loss of Sim-GAMP and present good robustness in mismatched cases. Compared with state-of-the-art (SOA) MIMO SC-FDMA detectors, the proposed DL-GAMP enjoys advances in both BER performance (up to $\mathbf {0.83}$ dB gain at BER $=\mathbf {10^{-3}}$ ) and computational complexity ( $\mathbf {66.44}$ % reduction at most).
- 15 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (7) , 10767-10771
A Deep-Learning-Aided Message Passing Detector for MIMO SC-FDMA
In this paper, a multiple-input multiple-output (MIMO) single carrier frequency division multiple access (SC-FDMA) detector is proposed, named simplified generalized approximate message passing (Sim-GAMP), which is derived from the existing low-complexity GAMP (LC-GAMP). To improve its BER performance, Sim-GAMP is further developed by means of deep learning (DL) techniques, which is called DL-GAMP in this work. Specifically, the layers of DL-GAMP are constructed by unfolding iterations of Sim-GAMP and introducing learnable parameters, thus contributing to better BER performance. Numerical results demonstrate that DL-GAMP can mitigate the performance loss of Sim-GAMP and present good robustness in mismatched cases. Compared with state-of-the-art (SOA) MIMO SC-FDMA detectors, the proposed DL-GAMP enjoys advances in both BER performance (up to $\mathbf {0.83}$ dB gain at BER $=\mathbf {10^{-3}}$ ) and computational complexity ( $\mathbf {66.44}$ % reduction at most).
- 31 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (4) , 4801-4813
Approximate Message Passing MIMO Detector With Frequency Domain Initialization for SC-FDMA
In this paper, multiple-input multiple-output (MIMO) single-carrier frequency division multiple access (SC-FDMA) detection algorithms, including low-complexity approximate message passing (LC-AMP) and frequency domain initialization (FDI) benefitted LC-AMP (FDI-LC-AMP), are proposed. The detailed derivation of LC-AMP algorithm starting from approximate message passing (AMP) is provided. Adaptive nearest-neighbor approximation (ANNA) and modified message damping strategy are illustrated. Additionally, FDI is adopted to improve the convergence performance of LC-AMP. The simulation results show that the proposed LC-AMP/FDI-LC-AMP with ANNA exhibit superior performance compared to other state-of-the-art (SOA) MIMO SC-FDMA detectors with and without channel encoding. The trade-off between complexity and performance is also discussed. Benefitting from FDI and ANNA, FDI-LC-AMP with ANNA can obtain 22.9%-44.0% reduction in complexity compared to the SOAs while performing better.
- 31 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (4) , 4801-4813
Approximate Message Passing MIMO Detector With Frequency Domain Initialization for SC-FDMA
In this paper, multiple-input multiple-output (MIMO) single-carrier frequency division multiple access (SC-FDMA) detection algorithms, including low-complexity approximate message passing (LC-AMP) and frequency domain initialization (FDI) benefitted LC-AMP (FDI-LC-AMP), are proposed. The detailed derivation of LC-AMP algorithm starting from approximate message passing (AMP) is provided. Adaptive nearest-neighbor approximation (ANNA) and modified message damping strategy are illustrated. Additionally, FDI is adopted to improve the convergence performance of LC-AMP. The simulation results show that the proposed LC-AMP/FDI-LC-AMP with ANNA exhibit superior performance compared to other state-of-the-art (SOA) MIMO SC-FDMA detectors with and without channel encoding. The trade-off between complexity and performance is also discussed. Benefitting from FDI and ANNA, FDI-LC-AMP with ANNA can obtain 22.9%-44.0% reduction in complexity compared to the SOAs while performing better.