We propose an alternating optimization (AO) approach to enhance the performance of energy-efficient multi-user visible light communication (MU-VLC) systems, which integrates both lighting and communication functions using multi-input multi-output (MIMO) technology and sub-array architecture. The rate-splitting multiple access (RSMA) technique is used to fully exploit the VLC channel as well as flexibly handle the interference and noise. Numerical results exhibit the superiority of the proposed scheme over the minimum mean squared error (MMSE) and successive interference cancellation (SIC) methods under various user quality of service (QoS) requirements. Moreover, the energy efficiency (EE) achieved by the proposed scheme surpasses that of existing VLC MIMO systems, highlighting its potential for green communication applications.

The rapid proliferation of Internet of Things (IoT) is generating an unprecedented volume of distributed data, necessitating efficient decentralized learning paradigms. Federated learning (FL) has emerged as a compelling distributed collaborative intelligence framework, renowned for its privacy protection benefits. However, the communication overhead associated with intermediate model exchanges remains a critical bottleneck in FL. Aiming at reducing the communication cost of FL equipped with promising over-the-air computation (AirComp) technique, this work designs specialized model compression schemes for both uplink and downlink communications. For uplink transmission with AirComp, we analyze its unique constraints and propose a hybrid global sparsification scheme that combines the benefits of conventional Top-k and Rand-k algorithms. We further develop an algorithm to strategically allocate transmission budgets between the two concatenated sparsification operations, accounting for both model temporal correlation and the cost of index synchronization. For downlink transmission, we introduce a group-based mixed-precision quantization (MPQ) scheme and integrates the broadcast of grouping information with uplink sparsification pattern to further mitigate communication burden. Moreover, we conduct theoretical analysis under realistic channel conditions and typical FL settings to validate the advantages and establish convergence guarantees of our approaches. Experimental results demonstrate that, compared to existing schemes, the proposed methods significantly improve communication efficiency and ensure client scalability, and concurrently verify the benefits of the uplink-downlink synergistic design.

In data driven deep learning, distributed sensing and joint computing bring heavy load for computing and communication. To face the challenge, over-the-air computation (OAC) has been proposed for multi-sensor data aggregation, which enables the server to receive a desired function of massive sensing data during communication. However, the strict synchronization and accurate channel estimation constraints in OAC are hard to be satisfied in practice, leading to time and channel-gain misalignment. The paper formulates the misalignment problem as a non-blind image deblurring problem. At the receiver side, we first use the Wiener filter to deblur, followed by a U-Net network designed for further denoising. Our method is capable to exploit the inherent correlations in the signal data via learning, thus outperforms traditional methods in term of accuracy. Our code is available at https://github.com/auto-Dog/MOAC_deep.

Federated learning (FL) is a promising solution to harness the advances of machine learning under the premise of privacy security, whereas the communication overhead of model exchange remains an obstacle to deploying FL in wireless networks. To tackle this challenge, we consider the non-uniform quantization of the global model in this work. By formulating the optimization of quantization intervals as a Markov decision process (MDP), we propose a deep reinforcement learning (DRL)- based approach to improve the performance of the quantizer for FL. Through crafting a compound reward function, the DRL agent is guided to reduce the quantization error and training loss simultaneously. Furthermore, a dual time-scale mechanism between FL and DRL is adopted to ensure that the actor and critic models of DRL converge more steadily. Simulations on various real-world datasets reveal that the proposed method can provide higher accuracy and faster convergence than the existing uniform quantizers, and can retain these benefits when applying the learned policy to a similar learning task.

Abstract Communication has been recognized as one of the primary challenges of federated learning (FL), but the actual communication algorithm or protocol design is still rarely involved in the existing studies. In the paper, viewing the model exchange in FL as a special kind of traffic, an unequal error protection (UEP) scheme is designed based on multi‐rate channel coding and multi‐layer modulation for it. To answer the question of how to make error control for FL when the wireless channel is no longer simplified as a pipeline, this paper firstly theoretically analyzes the impact of transmission error on machine leanring (ML) model, which reveals that the dynamic range of the weights should be taken into consideration. Guided by the analysis, the UEP scheme is applied to FL in multiple perspectives including parameter, network and time. Furthermore, a UEP‐based adaptive coding method is developed for the case with dynamic signal‐to‐noise ratio (SNR) to ensure faster and more stable convergence of the FL model while saving as much bandwidth as possible. Comprehensive numerical simulation on several real‐world datasets verifies that the proposed UEP transmission schemes can indeed bring significant benefits in accuracy, robustness and efficiency, especially when the channel condition is poor.

In this paper, we study the efficient communication design, including both uplink and downlink communications, for wireless federated learning (FL). We answer the question of what and how to communicate between clients and the parameter server and evaluate the impact of the various quantization and transmission options of the updated model on the learning performance. We provide new convergence analysis of the well-known FEDAVG under non-i.i.d. dataset distributions, partial clients participation, and finite-precision quantization in uplink and downlink communications. These analyses reveal that, in order to achieve an $\mathcal{O}(1/T)$ convergence rate with quantization, transmitting the weight requires increasing the quantization level at a logarithmic rate, while transmitting the weight differential can keep a constant quantization level. Comprehensive numerical evaluation on various real-world datasets reveals that the benefit of a FL-tailored uplink and downlink communication design is enormous – a carefully designed 1-bit quantization (3.1% of the floating-point baseline bandwidth) achieves 99.8% of the floating-point baseline accuracy at almost the same convergence rate on MNIST, representing the best known bandwidth-accuracy tradeoff to the best of the authors’ knowledge.

Communication has been known to be one of the primary bottlenecks of federated learning (FL), and yet existing studies have not addressed the efficient communication design, particularly in wireless FL where both uplink and downlink communications have to be considered. In this paper, we focus on the design and analysis of physical layer quantization and transmission methods for wireless FL. We answer the question of what and how to communicate between clients and the parameter server and evaluate the impact of the various quantization and transmission options of the updated model on the learning performance. We provide new convergence analysis of the well-known FED AVG under non-i.i.d. dataset distributions, partial clients participation, and finite-precision quantization in uplink and downlink communications. These analyses reveal that, in order to achieve anO(1/T) convergence rate with quantization, transmitting the weight requires increasing the quantization level at a logarithmic rate, while transmitting the weight differential can keep a constant quantization level. Comprehensive numerical evaluation on various real-world datasets reveals that the benefit of a FL-tailored uplink and downlink communication design is enormous - a carefully designed quantization and transmission achieves more than 98% of the floating-point baseline accuracy with fewer than 10% of the baseline bandwidth, for majority of the experiments on both i.i.d. and non-i.i.d. datasets. In particular, 1-bit quantization (3.1% of the floating-point baseline bandwidth) achieves 99.8% of the floating-point baseline accuracy at almost the same convergence rate on MNIST, representing the best known bandwidth-accuracy tradeoff to the best of the authors' knowledge.