Initial access (IA) is a procedure of establishing an initial connection between the base station (BS) and the users. In the fifth generation (5G) mobile communication system, the IA procedure includes beam management, which determines the beam pairs for random access (RA) and data transmission by beam sweeping. The existing beam sweeping method in the 3-rd generation partnership project (3GPP) standard mainly uses a predefined uniform beamforming codebook and sweeps the beams progressively, which is time-consuming and highly inflexible. In this paper, inspired by the fact that the highly non-uniform environment and user distribution mean part of the beam sweeping might be less beneficial, we propose a novel learning-based IA framework for the BS to optimize the beam sweeping patterns. Specifically, we resort to the deep reinforcement learning (DRL) approach to implicitly obtain the unknown environment and user distribution properties by continuously interacting with the environment, and then make decisions based on the rewards achieved by past actions. The simulation results show that our proposed scheme can save much time compared with the new radio (NR) and optimization methods under different datasets and conditions, which greatly improves the beam sweeping efficiency.

Conventional wireless communication receivers guided by Bayesian inference methods need to know the exact statistical relationship among variables, which is hard to obtain accurately in wireless contexts, thus limiting the system performance. The recently emerging artificial intelligence (AI)-empowered algorithms have shown striking performances in exploring the implicit relationship among variables with specially designed Neural Networks (NNs). Therefore, it is preferable to integrate NNs with BPs in receiver design. Such approaches also leverage NNs’ lack of reasoning ability in large state spaces and traditional BPs’ lack of reasoning depth. However, conventional receiver modules are usually designed based on explicit mathematical derivations, which cannot be easily substituted with data-driven NNs as they may break the overall inner relationship of the algorithm. In this paper, we investigate how to beneficially incorporate NNs into the existing Belief Propagation (BP)-based framework, taking the traditional semi-blind estimation problem in an Orthogonal Frequency-Division Multiplexing (OFDM) receiver as an example. Unlike existing deep-unfolding approaches, we simply utilize NNs as embedded functional units rather than duplicate denoising modules. Through qualitative discussions and numerical results, we illustrate the characteristics, principles, and differences of our proposed architecture compared to the traditional BP framework and show the dramatic performance improvements brought by incorporating NNs with BP in this well-investigated problem. Recalling that the state evolution of NNs is different from that of traditional BP methods, we give some new insights and design principles which are somehow counterfactual. We also raise some open issues on the incorporated framework.

This paper investigates semantic energy efficiency in a non-orthogonal multiple access (NOMA) enabled integrated sensing and semantic communication (ISSC) system. The model involves the base station (BS) transmitting information to multiple users while performing target sensing using dedicated beamforming. In the considered model, the BS needs to transmit substantial text data to each user using text semantic communication techniques while sensing the targets with certain constraints. Our goal is to maximize semantic energy efficiency and meet semantic communication and sensing accuracy requirements. We formulate an optimization problem for the beamforming matrix and semantic parameter, employing the Dinkelbach's algorithm for simplification and proposing an iterative solution. Numerical results validate the efficacy of the NOMA-ISSC scheme.

In massive Multiple-Input Multiple-Output (MIMO) -Orthogonal Frequency Division Multiplexing (OFDM) systems, channel estimation incurs high pilot overhead due to the high channel dimension, prompting the exploration of various algorithms to mitigate this cost. Semi-blind estimation, which usually employs a traditional iterative algorithm based on Bayesian infer-ence, proves effective in enhancing estimation performance with a limited number of pilots. Meanwhile, Neural Network (NN)-based channel mapping and prediction methods have demonstrated potential in predicting the full channel matrix with the estimation of a proportion, reducing the pilot overhead. However, how to merge these two methods for less pilot overhead is yet to be investigated. This paper introduces a hybrid model and data driven method that combines semi-blind estimation with NN-based frequency domain channel mapping, leveraging data priors as the inference algorithm while harnessing the nonlinear mapping capability offered by NNs. The proposed architecture holds promise for extension to other applications where the synergistic combination of inference algorithms and NN s proves advantageous. Numerical results validate the effectiveness of the proposed algorithm.

Recently , over-the-air computation (OAC) has shown potential in realizing computation tasks over wireless transmission. Through proper transmit and receive beamforming design, multiple-input multiple-output (MIMO)-based OAC systems can even realize partial functions of neural networks (NNs). In this paper, we propose an OAC-NN with reconfigurable intelligent surface (RIS)-aided MIMO, in which the NN computation task can be realized through updating the RIS reflection matrix. In the proposed structure, the communication system can complete the overall simple NN-based tasks only through multiple rounds of transmissions without introducing any additional computing resources. Numerical results reflect the effectiveness of the proposed scheme and the tradeoff between communication costs and computing performance.

In split machine learning (ML), different partitions of a neural network (NN) are executed by different computing nodes, requiring a large amount of communication cost. As over-the-air computation (OAC) can efficiently implement all or part of the computation at the same time of communication, thus by substituting the wireless transmission in the traditional split ML framework with OAC, the communication load can be eased. In this paper, we propose to deploy split ML in a wireless multiple-input multiple-output (MIMO) communication network utilizing the intricate interplay between MIMO-based OAC and NN. The basic procedure of the OAC split ML system is first provided, and we show that the inter-layer connection in a NN of any size can be mathematically decomposed into a set of linear precoding and combining transformations over a MIMO channel carrying out multi-stream analog communication. The precoding and combining matrices which are regarded as trainable parameters, and the MIMO channel matrix, which are regarded as unknown (implicit) parameters, jointly serve as a fully connected layer of the NN. Most interestingly, the channel estimation procedure can be eliminated by exploiting the MIMO channel reciprocity of the forward and backward propagation, thus greatly saving the system costs and/or further improving its overall efficiency. The generalization of the proposed scheme to the conventional NNs is also introduced, i.e., the widely used convolutional NNs. We demonstrate its effectiveness under both the static and quasi-static memory channel conditions with comprehensive simulations.

Federated learning (FL) is a privacy-preserving machine learning setting that enables many devices to jointly train a shared global model without the need to reveal their data to a central server. However, FL involves a frequent exchange of the parameters between all the clients and the server that coordinates the training. This introduces extensive communication overhead, which can be a major bottleneck in FL with limited communication links. In this paper, we consider training the binary neural networks (BNNs) in the FL setting instead of the typical real-valued neural networks to fulfill the stringent delay and efficiency requirement in wireless edge networks. We introduce a novel FL framework of training BNNs, where the clients only upload the binary parameters to the server. We also propose a novel parameter updating scheme based on the Maximum Likelihood (ML) estimation that preserves the performance of the BNN even without the availability of aggregated real-valued auxiliary parameters that are usually needed during the training of the BNN. Moreover, for the first time in the literature, we theoretically derive the conditions under which the training of BNN is converging. Numerical results show that the proposed FL framework significantly reduces the communication cost compared to the conventional neural networks with typical real-valued parameters, and the performance loss incurred by the binarization can be further compensated by a hybrid method.