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.

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.

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.

Many convolutional neural networks (CNNs) for single image deblurring employ a U-Net structure to estimate latent sharp images. Having long been proven to be effective in image restoration tasks, a single lane of encoder-decoder architecture overlooks the characteristic of deblurring, where a blurry image is generated from complicated blur kernels caused by tangled motions. Toward an effective network architecture for single image deblurring, we present complemental sub-solution learning with a one-encoder-two-decoder architecture. Observing that multiple decoders successfully learn to decompose encoded feature information into directional components, we further improve both the network efficiency and the deblurring performance by rotating and sharing kernels exploited in the decoders, which prevents the decoders from separating unnecessary components such as color shift. As a result, our proposed network shows superior results compared to U-Net while preserving the network parameters, and using the proposed network as the base network can improve the performance of existing state-of-the-art deblurring networks.

Coarse-to-fine strategies have been extensively used for the architecture design of single image deblurring networks. Conventional methods typically stack sub-networks with multi-scale input images and gradually improve sharpness of images from the bottom sub-network to the top sub-network, yielding inevitably high computational costs. Toward a fast and accurate deblurring network design, we revisit the coarse-to-fine strategy and present a multi-input multi-output U-net (MIMO-UNet). The MIMO-UNet has three distinct features. First, the single encoder of the MIMO-UNet takes multi-scale input images to ease the difficulty of training. Second, the single decoder of the MIMO-UNet outputs multiple deblurred images with different scales to mimic multi-cascaded U-nets using a single U-shaped network. Last, asymmetric feature fusion is introduced to merge multi-scale features in an efficient manner. Extensive experiments on the GoPro and RealBlur datasets demonstrate that the proposed network outperforms the state-of-the-art methods in terms of both accuracy and computational complexity. Source code is available for research purposes at https://github.com/chosj95/MIMO-UNet.

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.

This paper proposes a compressed sensing-based random access protocol (CS-RACH), which is suitable for servicing a large number of machine-type communication devices in Internet of Things (IoT) network. In CS-RACH, we utilize a larger number of unique preambles compared to conventional LTE-RACH, however, the compressed sensing technique makes it possible to simultaneously detect the users with high accuracy. Compared to the user detection in conventional LTE-RACH, the proposed user detection can get rid of preamble collisions and decrease the collision probability, thereby the overall access latency is significantly reduced. To prove the benefits of the proposed CS-RACH, we mathematically analyze and compare access latency performance of LTE-RACH and CS-RACH. In particular, based on the least absolute shrinkage and selection operator approach, we derive a normalized throughput, access success probability, and average access latency. Our simulation results also exhibit that the proposed CS-RACH considerably reduces the access latency under reasonable conditions in IoT environments.

In this paper, we consider user scheduling and power allocation techniques for downlinks where transmitted message must be kept confidential from all other unintended users and the number of users is larger than that of access point (AP) antennas. In order to maximize network secrecy rate with a limited number of antennas, the AP schedules a user who can achieve the largest secrecy rate and exploits multiple antennas for transmitting artificial noise to the unintended users. We derive the closed-form expressions of the optimal power allocation for the artificial noise and the average network secrecy rate through asymptotic analysis. The analysis results provide the insightful information on the behaviors of the user scheduling and the effect of system parameters on the network secrecy rate performance. It is shown that opportunistic scheduling with artificial noise transmission technique is more effective than the conventional precoding techniques for multiple confidential message transmissions in terms of network secrecy rate. Furthermore, we have also discussed a transmission strategy for the worst case scenarios where all unintended users are treated as colluding eavesdroppers. The simulations confirm the validity of the analysis results and provide additional information on the network performance in general cases.

This paper considers a two-way relay network, in which two sources exchange data through a relay and a cooperative jammer transmits an artificial noise (AN) while a number of nearby eavesdroppers overhear to recover data from both sources. The relay harvests energy from the two source signals and the AN, and then, uses this harvested energy to forward the received signals to the two sources. Each source eliminates its own signal from the relaying signal by self-cancellation and then decodes the data signal received from the other source. For this wireless-powered two-way relay system, we propose two secure relay protocols based on power splitting and time switching techniques. The two protocols are power splitting-based two-way relaying (PS-TWR) and time switching-based two-way relaying (TS-TWR), in which the relay, respectively, controls the power splitting ratio (p) and time switching ratio (α), in order to achieve a balance between the data receiving and the energy harvesting. The optimal values of p and α for each protocol are found analytically to maximize the minimum guaranteed secrecy capacity (CSmin) considering multiple eavesdroppers in high signal-to-noise ratio environments. Numerical results show that both the PS-TWR and TS-TWR protocols using the optimized values of p and α achieve the near-optimal CSminno matter how many eavesdroppers exist anywhere. Comparisons of the two protocols in various scenarios also show that PS-TWR achieves better CSminthan TS-TWR because PS-TWR inherently has a shorter vulnerable time for eavesdropping than TS-TWR.

A two-hop relay network is considered, in which an eavesdropper can overhear the relaying signal. To prevent the eavesdropper from decoding this signal, a destination transmits a jamming noise while a source transmits the data signal to the relay. At the same time, the relay can harvest energy from both the source signal and the jamming noise, and use this harvested energy to forward the received signal to the destination. In such a wireless-powered relay system with cooperative jamming, we propose two adaptive relaying schemes based on power splitting and time switching techniques. In the proposed power splitting-based relaying (PSR) and time switching-based relaying (TSR) schemes, the relay controls the power splitting ratio (p) and time switching ratio (α), respectively, in order to achieve a balance between signal processing and energy harvesting. We find analytically the optimal values of p and α in each scheme to maximize the secrecy capacity under the assumption of high signal-to-noise ratio (SNR). Interestingly, although the eavesdropper's channel state information (CSI) is used in the derivation of the optimal control parameters (p and α), they are shown not to be affected by the eavesdropper's CSI in a high SNR regime. This implies that the proposed schemes can be effective even for practical environments where there is no eavesdropper's CSI. Furthermore, simulation results show that they well coincide with the exact solutions in practical environments even though the closed-form solutions are obtained with a high SNR assumption. Moreover, the comparisons of PSR and TSR in various scenarios show that the two relaying schemes have complementary performances depending on the network conditions. Specifically, PSR achieves greater secrecy capacity than TSR when the channel condition is unfavorable to the eavesdropper for wiretapping.

We present a novel approach for improving the shape statistics of medical image objects by generating correspondence of skeletal points. Each object's interior is modeled by an s-rep, i.e., by a sampled, folded, two-sided skeletal sheet with spoke vectors proceeding from the skeletal sheet to the boundary. The skeleton is divided into three parts: the up side, the down side, and the fold curve. The spokes on each part are treated separately and, using spoke interpolation, are shifted along that skeleton in each training sample so as to tighten the probability distribution on those spokes' geometric properties while sampling the object interior regularly. As with the surface/boundary-based correspondence method of Cates et al., entropy is used to measure both the probability distribution tightness and the sampling regularity, here of the spokes' geometric properties. Evaluation on synthetic and real world lateral ventricle and hippocampus data sets demonstrate improvement in the performance of statistics using the resulting probability distributions. This improvement is greater than that achieved by an entropy-based correspondence method on the boundary points.

From the perspective of a single user, occupying all available radio access technologies (RAT) is always beneficial for increasing data rate. However, due to the interactions among the users, occupying multiple RATs at each user might not be optimal in a network view point. This paper studies the optimal association strategy to maximize the system throughput in multi-RAT cellular networks. To this end, with stochastic geometry, we derive the network sum rates by considering two association modes; single RAT association (altruistic association) and multiple RATs association (selfish association). By comparing the network sum rates for the two association modes, we show that optimally balancing the number of users between two association modes is necessary to improve the performances. With some numerical examples, we investigate how various system parameters, such as the number of carrier components, user density, and density of access points, etc, affect on the optimal portion of the users between two association modes.

From the perspective of a single user, concurrent utilization of all available radio access technologies (RATs) is always beneficial for increasing individual data rate. However, given the interactions among users, occupying multiple RATs at each user might not be optimal from a network-wide viewpoint. In this paper, we explore the answer to the posed question what is the optimal access strategy in multi-RAT cellular networks, solving a distributed RAT access control problem for maximizing network throughput. With stochastic geometry, we analytically evaluate network throughputs for two different access modes: 1) single RAT access (altruistic access) and 2) simultaneous multiple RATs access (selfish access). Comparing the network throughputs for the two modes, we first show that the network throughput can be maximized by properly mixing the two access modes and then derive the optimal portions of each mode in a network, which motivates a distributed RAT access control in a probabilistic sense. The optimal mixture of the two modes controls scheduling contention and interference among users to maximize the network throughput. We also analyze the effects of various system parameters, such as the number of frequency sub-bands for each RAT, user density, and access point density, on the optimal portions of the two access modes.

Each video clip has different degrees of popularity and a user can cache some popular videos in preparation for future requests. The cached data work as a buffer against the playback delay due to unstable wireless channels. Since the storage space for caching is limited while the number of video clips is tremendously large, a memory efficient caching strategy is important for maximizing the caching gain. We propose a user prefix caching scheme in a downlink network where each user caches an initial part (i.e., prefix) of each popular video to minimize the average playback delay. We derive closed-form expressions of the optimal prefix size and the corresponding average playback delay. Our results show the effects of system parameters on the average playback delay and provide an insightful guideline on memory-efficient caching. Moreover, in the scenarios where a constraint on tolerable delay is given, we derive the minimum storage space required at each user and the maximum number of supportable users at a base station in wireless caching networks, which provide a useful insight on designing wireless caching networks.

This paper considers random multiple access in a network where only a small portion of users have data to forward and transmit packets in each time slot because the user activity ratio is not high in practice. For this reason, the access point (AP) has to not only identify the users who transmitted but also decode the received data codewords. Exploiting the sparsity of transmitting users, Lasso, which is a well-known practical compressed sensing algorithm, is applied for efficient user identification. The compressed sensing algorithm enables the AP to handle more users than the conventional random multiple access schemes do. We develop distributed scheduling methods for maximizing the system sum throughput, and we analyze the corresponding optimal throughput for three different cases of channel knowledge, i.e., the channel state information at the transmitter (CSIT), the channel state information at the receiver (CSIR), and the imperfect channel state information at the receiver (ImCSIR). We also derive the closed-form expressions of asymptotically optimal scheduling parameters and the corresponding maximum sum throughput for each CSI assumption. The results show the effects of system parameters on the sum throughput and provide useful insights on using compressed sensing for throughput maximization in random multiple access schemes.

Cognitive radio (CR) systems are typically classified into two types. In overlay CR, a secondary user opportunistically accesses primary user's spectrum when it is unused, while a secondary user in underlay CR accesses the spectrum at any time as long as the interference power received at a primary user is below a certain level. This paper investigates a hybrid CR system where a secondary user probabilistically changes its spectrum access mode for secondary user's throughput maximization while guaranteeing primary user's target throughput. Basically, the hybrid CR works like overlay CR. The secondary user constantly monitors activity of the primary user and transmits data with maximum transmit power if transmission of the primary user is not detected. However, the only difference from overlay CR is that even when transmission of the primary user is detected, the secondary users are able to access the spectrum with regulated low transmit power with switching rate c. We show that there is a trade-off of increasing c due to recursive interactions between primary and secondary users, and optimizing c is crucial to balance the gain and loss of secondary user's underlay spectrum access. The proposed hybrid CR is shown to achieve higher throughput and be more robust to detection errors than conventional CR systems.

We investigate the cooperative diversity gain in underlay cognitive radio (CR) systems where secondary users access the spectrum licensed to primary users under an interference power constraint at primary receivers. Our analysis shows that the cooperative diversity gain has different characteristics from that in non-CR systems since the transmit power of a secondary transmitter is limited by the interference power constraint. If a fixed interference power constraint is imposed, the cooperative diversity gain is lost regardless of the number of relays even if peak transmit power grows to infinity. We analyze the critical value of peak transmit power beyond which increasing peak transmit power does not help in improving the outage probability. However, if a proportional interference power constraint is imposed, a full diversity gain is shown to be attained even without instantaneous interference channel information at the secondary transmitters.

This paper analyzes the multiuser diversity gain in a cognitive radio (CR) system where secondary transmitters opportunistically utilize the spectrum licensed to primary users only when it is not occupied by the primary users. To protect the primary users from the interference caused by the missed detection of primary transmissions in the secondary network, minimum average throughput of the primary network is guaranteed by transmit power control at the secondary transmitters. The traffic dynamics of a primary network are also considered in our analysis. We derive the average achievable capacity of the secondary network and analyze its asymptotic behaviors to characterize the multiuser diversity gains in the CR system.

This paper proposes a novel generalized selection combining (GSC)/maximal ratio combining(MRC) switching technique based on error detection in cooperative communications with amplify-and-forward (AF) relays. In order to reduce the excessive burden of MRC with all diversity paths at the destination node, the destination node decides if it performs GSC with order N(< K) or MRC with order K + 1 based on the error detection, where K is the number of relay nodes. Our analytical and simulation results show that the proposed GSC/MRC switching outperforms the conventional MRC in terms of outage probability in AF based cooperative communications since the proposed scheme effectively reduces the spectral efficiency loss with the help of error detection codes.