Beamforming is an effective means to improve the quality of the received signals in multiuser multiple-input-singleoutput (MISO) systems. Traditionally, finding the optimal beamforming solution relies on iterative algorithms, which introduces high computational delay and is thus not suitable for realtime implementation. In this paper, we propose a deep learning framework for the optimization of downlink beamforming. In particular, the solution is obtained based on convolutional neural networks and exploitation of expert knowledge, such as the uplink-downlink duality and the known structure of optimal solutions. Using this framework, we construct three beamforming neural networks (BNNs) for three typical optimization problems, i.e., the signal-to-interference-plus-noise ratio (SINR) balancing problem, the power minimization problem, and the sum rate maximization problem. For the former two problems the BNNs adopt the supervised learning approach, while for the sum rate maximization problem a hybrid method of supervised and unsupervised learning is employed. Simulation results show that the BNNs can achieve near-optimal solutions to the SINR balancing and power minimization problems, and a performance close to that of the weighted minimum mean squared error algorithm for the sum rate maximization problem, while in all cases enjoy significantly reduced computational complexity. In summary, this work paves the way for fast realization of optimal beamforming in multiuser MISO systems.

By exploiting the computing power and local data of distributed clients, federated learning (FL) features ubiquitous properties such as reduction of communication overhead and preserving data privacy. In each communication round of FL, the clients update local models based on their own data and upload their local updates via wireless channels. However, latency caused by hundreds to thousands of communication rounds remains a bottleneck in FL. To minimize the training latency, this work provides a multi-armed bandit-based framework for online client scheduling (CS) in FL without knowing wireless channel state information and statistical characteristics of clients. Firstly, we propose a CS algorithm based on the upper confidence bound policy (CS-UCB) for ideal scenarios where local datasets of clients are independent and identically distributed (i.i.d.) and balanced. An upper bound of the expected performance regret of the proposed CS-UCB algorithm is provided, which indicates that the regret grows logarithmically over communication rounds. Then, to address non-ideal scenarios with non-i.i.d. and unbalanced properties of local datasets and varying availability of clients, we further propose a CS algorithm based on the UCB policy and virtual queue technique (CS-UCB-Q). An upper bound is also derived, which shows that the expected performance regret of the proposed CS-UCB-Q algorithm can have a sub-linear growth over communication rounds under certain conditions. Besides, the convergence performance of FL training is also analyzed. Finally, simulation results validate the efficiency of the proposed algorithms.

Beamforming is considered one of the most important techniques for designing advanced multiple-input and multiple-output (MIMO) systems. Among these design criterions, sum rate maximization (SRM) under a total power constraint is a challenge issue due to its nonconvexity. Existing techniques for the SRM problem only obtain suboptimal solutions but require huge amount of computation due to their complex matrix operation and iterations. Unlike these conventional methods, we propose a deep learning based fast beamforming design method without complex matrix operation and iterations. Specifically, we first derive a heuristic solution structure of the downlink beamforming through the virtual equivalent uplink channel based on optimum MMSE receiver which separates the problem into power allocation and virtual uplink beamforming (VUB) design. Next, beamforming prediction network (BPNet) is designed to perform the joint optimization of power allocation and VUB. Moreover, the BPNet is trained offline using two-step training strategy. Simulation results demonstrate that our proposed method is fast while obtains the comparable performance to the state-of-the-art method.

In the downlink transmission scenario, power allocation and beamforming design at the transmitter are essential when using multiple antenna arrays. This paper considers a multiple input-multiple output broadcast channel to maximize the weighted sum-rate under the total power constraint. The classical weighted minimum mean-square error (WMMSE) algorithm can obtain suboptimal solutions but involves high computational complexity. To reduce this complexity, we propose a fast beamforming design method using unsupervised learning, which trains the deep neural network (DNN) offline and provides real-time service online only with simple neural network operations. The training process is based on an end-to-end method without labeled samples avoiding the complicated process of obtaining labels. Moreover, we use the ’APoZ’-based pruning algorithm to compress the network volume, which further reduces the computational complexity and volume of the DNN, making it more suitable for low computation-capacity devices. Finally, experimental results demonstrate that the proposed method improves computational speed significantly with performance close to the WMMSE algorithm.

By offloading tasks from the mobile device (MD) to its nearby deployed access points (APs), each of which is connected to one server for task processing, computation offloading can strike a balance between MD’s task execution delay and energy consumption in mobile edge computing (MEC) systems. Considering communication and computation dynamics in MEC systems, we aim to design online computation offloading mechanisms in this paper to minimize the time average expected task execution delay under the constraint of average energy consumption. Firstly, with known current channel gains between the MD and APs as well as available computing capability at MEC servers, we leverage the Lyapunov optimization framework to make an optimal one-slot decision on MD’s transmit power allocation and MEC server selection. On this basis, we then consider a more realistic scenario, where it is difficult to capture current available computing capability at MEC servers, and combine the multi-armed bandit framework for an online learning based MEC server selection algorithm. Finally, through theoretical analyses and extensive simulations, we demonstrate the near-optimality and feasibility of our proposed algorithms, and present that our proposed algorithms fully explore the interplay between communication and computation with enriched user experience and reduced energy consumption.

Federated learning (FL) is promising in enabling large-scale model training by massive devices without exposing their local datasets. However, due to limited wireless resources, traditional cloud-based FL system suffers from the bottleneck of communication overhead in core network. Fueled by this issue, we consider a hierarchical FL system and formulate a joint problem of edge aggregation interval control and resource allocation to minimize the weighted sum of training loss and training latency. To quantify the learning performance, an upper bound of the average global gradient deviation, in terms of the edge aggregation interval, the training latency, and the number of successfully participating devices, is derived. Then an alternative problem is formulated, which can be decoupled into an edge aggregation interval control problem and a resource allocation problem, and solved by an iterative optimization algorithm. Specifically, given the resource allocation strategy, a relaxation and rounding method is proposed to optimize the edge aggregation interval. The problem of resource allocation including training time allocation and bandwidth allocation is solved separately based on the convex optimization theory. Simulation results show that the proposed algorithm, compared to the benchmarks, can achieve higher learning performance with lower training latency, and is capable of adaptively adjusting the edge aggregation interval and the resource allocation strategy according to the training process.