The conventional FL methods face critical challenges in realistic wireless edge networks, where training data is both limited and heterogeneous, often leading to unstable training and poor generalization. To address these challenges in a principled manner, we propose a novel wireless FL framework grounded in Bayesian inference. By virtue of the Bayesian approach, our framework captures model uncertainty by maintaining distributions over local weights and performs distribution-level aggregation of local distributions into a global distribution. This mitigates local overfitting and client drift, thereby enabling more reliable inference. Nevertheless, adopting Bayesian FL increases communication overhead due to the need to transmit richer model information and fundamentally alters the aggregation process beyond simple averaging. As a result, conventional Over-the-Air Computation (AirComp), widely used to improve communication efficiency in standard FL, is no longer directly applicable. To overcome this limitation, we design a dedicated AirComp scheme tailored to Bayesian FL, which efficiently aggregates local posterior distributions at the distribution level by exploiting the superposition property of wireless channels. In addition, we derive an optimal transmit power control strategy, grounded in rigorous convergence analysis, to accelerate training under power constraints. Our analysis explicitly accounts for practical wireless impairments such as fading and noise, and provides theoretical guarantees for convergence. Extensive simulations validate the proposed framework, demonstrating significant improvements in test accuracy and calibration performance over conventional FL methods, particularly in data-scarce and heterogeneous environments.

Federated Learning (FL) enables training models across distributed devices while preserving data privacy by avoiding raw data sharing. However, it suffers from significant communication overhead. Event-Triggered FL (ETFL) addresses this issue by allowing devices to transmit updates only when substantial changes occur in the model. Nevertheless, this approach may result in imbalanced communication, where some devices communicate more frequently than others, leading to uneven model performance and slower overall convergence. To address this, we propose a new threshold-based method that dynamically adjusts each device’s communication frequency. Our method ensures balanced communication across devices and reduces the time required for each training iteration, ultimately accelerating convergence time. Furthermore, we analyze how a device’s communication affects the difference between its local model and the global model. Through extensive experiments, we demonstrate that the proposed method significantly reduces communication imbalance and achieves faster convergence compared to existing approaches. This result highlights the importance of balancing communication in federated learning to improve overall performance and ensure fairness across devices.

The surge in wireless devices and data traffic volume necessitates more efficient transmission methods. Multicasting has garnered consistent attention as a means to fulfill the increasing demand for more efficient data transmission methods. Nevertheless, leveraging multicast wireless networks for spatio-temporally asynchronous data requests poses challenges. In this context, this paper introduces a new multicast mechanism called set-up based merged multicast (SMMC) to minimize the delivery time of the requested file in wireless networks by considering the uncertainties inherent in wireless channels. The proposed mechanism comprises two phases. The first phase involves gathering asynchronous requests for a file from users experiencing diverse channel conditions. During this phase, packets of the requested file are transmitted individually in unicast mode within a specified set-up time. Following this, the second phase initiates multicast transmission, which sequentially handles the remaining packets of the file in multicast mode. In the proposed mechanism, we optimize the set-up time and transmission rates of both unicast and multicast modes to minimize the expected file delivery time by jointly taking into account the statistical characteristics of wireless channels, users' locations, and file popularity. Additionally, we also delve into a fine-tuned SMMC (FT-SMMC) by utilizing posterior information on the multicast group size and further improve the performance. Extensive simulations demonstrate that the proposed multicast transmission achieves substantial reductions in delivery time, and this performance gain is further enhanced by optimizing set-up time and transmission rates in diverse wireless network scenarios.

In this paper, we consider asynchronous federated learning (FL) over time-division multiple access (TDMA)-based communication networks. Considering TDMA for transmitting local updates can introduce significant delays to conventional synchronous FL, where all devices start local training from a common global model. In the proposed asynchronous FL approach, we partition devices into multiple TDMA groups, enabling simultaneous local computation and communication across different groups. This enhances time efficiency at the expense of staleness of local updates. We derive the relationship between the staleness of local updates and the size of the TDMA group in a training round. Moreover, our convergence analysis shows that although outdated local updates hinder appropriate global model updates, asynchronous FL over the TDMA channel converges even in the presence of data heterogeneity. Notably, the analysis identifies the impact of outdated local updates on convergence rate. Based on observations from our convergence rate, we refine asynchronous FL strategy by introducing an intentional delay in local training. This refinement accelerates the convergence by reducing the staleness of local updates. Our extensive simulation results demonstrate that asynchronous FL with the intentional delay can rapidly reduce global loss by lowering the staleness of local updates in resource-limited wireless communication networks.

Recent advances in deep learning (DL) have significantly enhanced automatic modulation classification (AMC), reducing the dependence on intricate feature engineering and substantially improving classification accuracy. Nonetheless, the conventional DL-based AMC methods have mostly adopted frequentist approaches, often exhibit limited adaptability and struggle to offer reliable uncertainty estimates for their predictions, especially when trained on limited data. To address these challenges, we propose a novel Bayesian DL framework tailored for the uncertainty-aware incremental AMC. Our framework involves a two-stage process, beginning with the optimization of model parameters using a cumulative loss function in frequentist approach, followed by the application of Bayesian Neural Network (BNN) techniques, such as Laplace approximation and variational inference, for statistical inference. This framework ensures robust calibration and scalability, enabling the classifier to incrementally refine its predictions with the reception of additional signal samples and to terminate the refinement process once the confidence score meets the predetermined classification accuracy target. In addition, these enhanced capabilities of our classifier can extend to identifying a new, unseen modulation scheme by discerning distinct confidence score patterns associated with varying sequence lengths. Simulation results validate the efficacy of the proposed uncertainty-aware incremental AMC by showing enhanced classification accuracy and superior calibration relative to conventional DL-based AMC methods in practical scenarios with limited training data.

This article proposes a deep learning-based power control method for maximizing the sum rate subject to rate requirements in the interference-limited device-to-device (D2D) communications. Based on the dynamic nature of D2D communications, we consider the environment where system parameters, such as the number of devices, rate requirements, and deployment area, unpredictably change over time. To deal with the low adaptability and scalability issues of the conventional deep learning-based approaches in dynamic environments, we develop an environment-adaptive power control method by leveraging graph neural network (GNN) architecture and meta-learning approach. In the developed method, we design the node feature and state update rule for GNN by taking into account the characteristics of power optimization problem and meta-train model by treating some past environments as meta-tasks. Simulation results demonstrate that the developed method outperforms the conventional GNN-based power control methods in terms of the average sum-throughput achievement ratio and adaptation speed to test environments.

In this article, we propose a deep reinforcement learning (DRL)-based geographic routing method designed to reduce retransmission delay in mobile wireless sensor networks (MWSNs) under disaster communication scenarios. Unstable wireless channels, which are common in these challenging environments, significantly contribute to communication delays. Unlike conventional approaches that assume error-free transmissions, our method accounts for wireless channel instability and node mobility. By incorporating key factors such as topological information and transmission stability, our approach optimizes routing decisions to minimize delays while maintaining packet delivery reliability. Simulation results demonstrate that our method outperforms conventional geographic routing algorithms, achieving lower retransmission delays and higher packet delivery ratios (PDRs) across various environments.

Mobile edge caching is regarded as a promising technology for reducing network latency and alleviating network congestion by efficiently offloading data traffic and computations to cache-enabled edge nodes. To fully leverage the benefits of edge caching, it is essential to jointly optimize caching and communication strategies, accounting for dynamic content request pattern and unstable nature of wireless mobile networks. Motivated by this, we study a joint cache replacement and user association strategy for minimizing the content delivery latency in cache-enabled cloud radio access network (C-RAN) where remote radio heads (RRHs) cache some contents for serving the content request without downloading the requested content from centralized baseband unit (BBU) via fronthaul. Unlike traditional cache placement strategies, our cache replacement facilitates gradual and timely updates while serving user content requests, without imposing additional network overhead. Specifically, whenever a user requests a content, BBU makes decisions on selecting a RRH for serving user request and on replacing the cached data of the selected RRH by taking into account the user location, cache status of RRHs, and impact on subsequent content deliveries. We optimize the selection of RRH to serve user request and the replacement of cached data by formulating a latency minimization problem using Markov Decision Process (MDP). This formulation considers the tradeoff between cache hit ratio and communication reliability. To develop an effective strategy for solving the MDP, we employ a deep reinforcement learning (DRL) algorithm and design a novel neural network structure and input feature map, specifically tailored to our problem domain. Simulation results show that the proposed approach learns effective strategy appropriate to a given environment, thereby outperforming not only the traditional rule-based strategies but also a typical DRL algorithm in terms of average latency. The proposed approach is shown to be relatively robust to time-variant content popularity by quickly adapting to new popularity distribution.

To tackle a Doppler sensitivity problem of orthogonal frequency division multiplexing (OFDM), orthogonal time frequency space (OTFS) has been investigated, where information is carried over delay-Doppler domain. In this paper, to improve communication reliability in doubly dispersive channel, an auto-encoder (AE)-based OTFS modulation and detection scheme is developed, where the transmit OTFS waveform and its associated detection scheme at the receiver are jointly optimized in a deep learning framework. However, the conventional AE architecture which takes one-hot encoded input vector is hard to be reused in OTFS due to its enormous input dimensionality that increases exponentially on the number of grid points in delay-Doppler domain. To overcome it, we divide the delay-Doppler grid into multiple subblocks and associate the one-hot encoded vector with each subblock. Then, by concatenating them, one multi-hot vector is formed and exploited as the input vector for the proposed AE-based OTFS modulation and detection. We also develop a meta-learning scheme to effectively train the AE-based OTFS transceiver for newly updated channel profile.

This paper proposes an over-the-air aggregation framework for federated learning (FL) in broadband wireless networks where not only edge devices but also a base station (BS) has its own local dataset. The proposed framework leverages the BS dataset to improve communication efficiency of FL by reducing the number of channel uses required for the model convergence as well as avoiding the signaling overhead incurred by power scale coordination among edge devices. We analyze the convergence to a stationary point without convexity assumption on the objective function. The analysis result reveals that the utilization of BS dataset improves the convergence rate and the update distortion caused by the limited power budget is a crucial factor hindering the model convergence. To facilitate the convergence, we develop an optimized power control method by solving the distortion minimization problem without assumptions on power scale coordination and global CSI at BS. Our simulation results validate that BS dataset is beneficial to reducing the number of channel uses for the model convergence and the developed power control method outperforms the conventional method in terms of both convergence rate and converged test accuracy. Furthermore, we identify some scenarios where the compression of local update can be helpful to reduce communication resources for model training.

Recently, federated learning (FL) has been receiving great attention as an effective machine learning method to avoid the security issue in raw data collection, as well as to distribute the computing load to edge devices. However, even though wireless communication is an essential component for implementing FL in edge networks, there have been few works that analyze the effect of wireless networks on FL. In this paper, we investigate FL in small-cell networks where multiple base stations (BSs) and users are located according to a homogeneous Poisson point process (PPP) with different densities. We comprehensively analyze the effects of geographic node deployment on the model aggregation in FL on the basis of stochastic geometry-based analysis. We derive the closed-form expressions of coverage probability with tractable approximations and discuss the minimum required BS density for achieving a target model aggregation rate in small-cell networks. Our analysis and simulation results provide insightful information for understanding the behaviors of FL in small-cell networks; these can be exploited as a guideline for designing the network facilitating wireless FL.

In conventional federated learning (FL), dataset at the parameter server (PS) is not usually considered but may enhance the performance of FL if available. The benefit of leveraging the dataset at PS can be multiplied in over-the-air aggregation-based FL where combating local gradient update distortion against channel fading is inordinately consequential. In this paper, an over-the-air aggregation framework for communication efficient FL is investigated in cache-enabled wireless edge networks where not only edge devices but also a base station (BS) has its own local dataset. The proposed framework leverages the BS dataset to reduce the number of channel uses necessary for the model convergence and to avoid the overhead incurred by power scale coordination and global channel state information (CSI) acquisition at BS. We present a sufficient condition for convergence to a stationary point without convexity assumption on the objective function. Based on the sufficient condition, a power control method is optimized to facilitate the model convergence without assumptions on power scale coordination and global CSI at BS. Our simulation results validate that BS dataset is beneficial to reduce the number of channel uses for the model convergence and the developed power control method outperforms the conventional method in terms of both convergence rate and converged test accuracy.

Thermoelectric technology, which has been receiving attention as a sustainable energy source, has limited applications because of its relatively low conversion efficiency. To broaden their application scope, thermoelectric materials require a high dimensionless figure of merit (ZT). Porous structuring of a thermoelectric material is a promising approach to enhance ZT by reducing its thermal conductivity. However, nanopores do not form in thermoelectric materials in a straightforward manner; impurities are also likely to be present in thermoelectric materials. Here, a simple but effective way to synthesize impurity-free nanoporous Bi0.4Sb1.6Te3 via the use of nanoporous raw powder, which is scalably formed by the selective dissolution of KCl after collision between Bi0.4Sb1.6Te3 and KCl powders, is proposed. This approach creates abundant nanopores, which effectively scatter phonons, thereby reducing the lattice thermal conductivity by 33% from 0.55 to 0.37 W m(-1) K-1. Benefitting from the optimized porous structure, porous Bi0.4Sb1.6Te3 achieves a high ZT of 1.41 in the temperature range of 333-373 K, and an excellent average ZT of 1.34 over a wide temperature range of 298-473 K. This study provides a facile and scalable method for developing high thermoelectric performance Bi2Te3-based alloys that can be further applied to other thermoelectric materials.

This paper presents a novel deep learning framework for autonomous clinical diagnosis by dealing with the training with poorly labeled clinical dataset. Partially labeled data and inconsistent labels from multiple annotators make the model hard to learn accurate diagnosis in frequently and drastically updated clinical dataset. Motivated by such difficulties, the proposed framework introduces the weighted combination of inconsistent labels by considering multiple annotators' expertise and adapt meta-learning approach for the quick adaptation to the updated dataset. Experimental results on the posterior pelvic tilt detection in a squat motion show the proposed approach outperforms the conventional learning approaches in terms of the convergence speed and the converged mean squared error.

With the recent advances in wireless radio access technologies (RATs), various types of RATs are coexisting in the current communication systems. This letter studies the distributed multi-radio access control for the random decentralized OFDMA-based multi-RAT wireless networks, where the randomly distributed multi-mode capable transmitters communicate with their designated receivers by probabilistically accessing to either multiple RATs or single RAT. We analyze the network throughputs and derive the optimal RAT access mode selection probability with stochastic geometry. We numerically evaluate how various system parameters affect on the optimal mode selection probability and validate that the optimal balance of two access modes can maximize the network throughput by efficiently controlling the network interference and aggregating the bandwidth.

In this paper, we consider wireless-powered secure communication with an energy harvesting receiver, which is allowed to harvest energy from the transmitted signals but not to decode information, and there is thus a requirement to keep the information secret from this potential eavesdropper. Considering co-channel interference among signal links, we find the optimal transmit power to maximize the sum rate of the signal links, while ensuring the requirements of information secrecy and energy harvesting. Due to the non-convexity of the optimization problem formulated here, we first derive suboptimal solutions using an iterative algorithm based on a dual method. In order to address the limitations caused by the use of the iterative algorithm, i.e., long computation time and suboptimality, we design an efficient deep neural network (DNN) framework and a novel training strategy as a means of combining supervised and unsupervised training. Specifically, the DNN is first pre-trained using labeled training data with the suboptimal solutions obtained from the iterative algorithm in a supervised manner; further training is then applied to the DNN using a well designed loss function in an unsupervised manner to enhance the training performance. Simulation results reveal that the proposed scheme achieves a near-optimal performance with a lower computation time than existing schemes. We also verify that the pre-training and the new loss function are effective in improving the speed of training of the DNN.

본 논문은 인체를 통신 매체로 활용하는 인체통신(human body communication, HBC) 환경에서 다수의 웨어러블 센서들의 정보를 데이터 컬렉터 또는 허브에 전송하는 다중접속 시스템을 고려한다. Point-to-point 인체통신시스템과 시분할 다중접속 (time division multiple access, TDMA) 방식이 접목된 기존의 인체통신 시스템과 달리 본 논문에서는 autoencoder 기반 딥러닝 학습 기법을 활용하여, 비직교 다중접속 (non-orthogonal multiple access, NOMA) 방식을 위한 전송 파형 최적화 및 다중 노드 검파 기법을 제안한다. 특히 인체통신 특성상 다중 노드들과 데이터 컬렉터와의 떨어진 거리에 따른 손실이 다른 환경에서 전송파형을 최적화하면서 노드들의 전력 최적화를 위한 신경망 구조를 제시한다. 또한 다중 노드 검파 성능 개선을 위한 NOMA 환경에서 간섭제거 학습에 용이한 신경망 구조를 제시하고자 한다. 모의 실험을 통해 인체 통신 환경에서의 제안한 autoencoder 기반 다중 노드검파 기법의 성능을 검증하였다.

In this paper, we propose novel resource allocation algorithms for ultra-reliable and low-latency communication (URLLC) in distributed antenna systems (DASs) and discuss the effectiveness of the receive antenna deployment and multiple access strategies in fulfilling the stringent reliability and latency requirements. We analyze achievable rates and reliabilities of the DASs with orthogonal multiple access (OMA) and non-orthogonal multiple access (NOMA) by taking into account the influence of the latency constraint and the non-identical channel distributions of the distributed antennas. Based on the performance analysis, we propose novel resource allocation algorithms for fulfilling the stringent performance requirement in OMA and NOMA. The numerical simulation results provide insightful information for understanding the impacts of multiple access strategy and antenna deployment on the spectral efficiency under the reliability and latency constraints. In particular, NOMA makes it possible to deal with more transmit nodes than OMA, and more dispersed antenna deployment facilitates to achieve a target URLLC performance with less spectrum resource as the ratio of transmit nodes to receive antennas increases. Consequently, the analysis and simulation results provide the meaningful information on the network design for mission-critical applications.

In this article, we consider a distributed antenna system to enable the downlink ultrareliable and low-latency communications (URLLC) for the industrial Internet of Things. To satisfy the low latency, the short packet length is exploited and accordingly, the achievable rate under the finite blocklength codes is considered in the analysis of the packet loss probability in downlink URLLC distributed antenna system. Specifically, the packet loss probability is approximated as a sum of the channel OP and the finite blocklength coding error probability. By using a stochastic geometry approach, we analytically derive the packet loss probability in terms of the system parameters (i.e., the densities of the DA ports and the receiving sensor nodes, the number of receive antennas, and the packet size/duration, etc.), when the sensor nodes and the DA ports are randomly distributed in a given area. From the analytic results, we show how to optimize the short blocklength coding error probability to minimize the overall packet loss probability. Furthermore, we also derive the required DA port density to satisfy the packet loss probability constraint for a given distributed antenna system parameters, which gives us a useful insight into the design of downlink distributed antenna system for URLLC.

In this paper, we propose a resource management method based on deep learning, which controls both the transmit power and the power splitting ratio to maximize the sum rate with low computational complexity in D2D networks with energy harvesting requirements. The introduction of the energy harvesting requirements to D2D networks makes it hard to design an effective resource management solution since the treatment of interference signals should be completely different from the conventional resource management focusing only on the rate maximization. To deal with drawbacks of the conventional deep learning-based approach, we propose a new training algorithm suitable for our resource management problem. Numerical simulations show that the proposed learning-based method outperforms the benchmark methods, which are derived from some relevant works, in most situations and achieves performances comparable to an exhaustive search in terms of the sum rate and energy outage probability. Although the conventional optimization-based method is derived to achieve the asymptotic optimal performance for a large network, the proposed deep learning method is shown to achieve almost the same performance with much lower computational complexity. Furthermore, simulation results offer new insights to the impact of the energy harvesting requirements on the behaviour of the optimal resource management.