Deep Unfolding with Kernel-based Quantization in MIMO Detection

2025-05-19

AI Summary

To address the accuracy degradation of deep-unfolding MIMO detection models (e.g., PGD-Nets, ADMM-Nets) under quantization on resource-constrained edge devices—caused by inaccurate activation distribution assumptions and static quantization step-size mismatches—this paper proposes a prior-free dynamic quantization method. First, kernel density estimation (KDE) and maximum mean discrepancy (MMD) are jointly employed to align inter-layer activation distributions, eliminating reliance on predefined assumptions (e.g., Gaussian or uniform). Second, a channel-adaptive dynamic step-size update mechanism is introduced, enabling fine-grained, input-aware trade-offs between detection accuracy and computational efficiency. Experiments demonstrate that the proposed method significantly improves quantized model accuracy while maintaining low inference latency, consistently outperforming conventional quantization-aware training (QAT) approaches across diverse channel conditions and bit-widths.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search
• Intelligent Robots: State Estimation

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The development of edge computing places critical demands on energy-efficient model deployment for multiple-input multiple-output (MIMO) detection tasks. Deploying deep unfolding models such as PGD-Nets and ADMM-Nets into resource-constrained edge devices using quantization methods is challenging. Existing quantization methods based on quantization aware training (QAT) suffer from performance degradation due to their reliance on parametric distribution assumption of activations and static quantization step sizes. To address these challenges, this paper proposes a novel kernel-based adaptive quantization (KAQ) framework for deep unfolding networks. By utilizing a joint kernel density estimation (KDE) and maximum mean discrepancy (MMD) approach to align activation distributions between full-precision and quantized models, the need for prior distribution assumptions is eliminated. Additionally, a dynamic step size updating method is introduced to adjust the quantization step size based on the channel conditions of wireless networks. Extensive simulations demonstrate that the accuracy of proposed KAQ framework outperforms traditional methods and successfully reduces the model's inference latency.

Problem

• Energy-efficient MIMO detection for edge computing deployment
• Overcoming performance loss in quantization for deep unfolding models
• Dynamic quantization step adjustment for varying wireless channel conditions

Innovation

• Kernel-based adaptive quantization for deep unfolding
• Joint KDE and MMD aligns activation distributions
• Dynamic step size updates based on channel conditions