Achievable Rate of Near-Field Communications Based on Physically Consistent Models

Abstract

Data-driven machine learning (ML) is promoted as one potential technology to be used in next-generations wireless systems. This led to a large body of research work that applies ML techniques to solve problems in different layers of the wireless transmission link. However, most of these applications rely on supervised learning which assumes that the source (training) and target (test) data are independent and identically distributed (i.i.d). This assumption is often violated in the real world due to domain or distribution shifts between the source and the target data. Thus, it is important to ensure that these algorithms generalize to out-of-distribution (OOD) data. In this context, domain generalization (DG) tackles the OODrelated issues by learning models on different and distinct source domains/datasets with generalization capabilities to unseen new domains without additional finetuning. Motivated by the importance of DG requirements for wireless applications, we present a comprehensive overview of the recent developments in DG and the different sources of domain shift. We also summarize the existing DG methods and review their applications in selected wireless communication problems, and conclude with insights and open questions.

Universal Function Approximation Through Over-the-Air Computing: A Deep Learning Approach

“…In practical implementations, alterations in environmental factors, e.g., channel distribution, number of users, would necessitate the retraining of the cDNN. The task of tackling the dynamic changes in the context of DNNs is left for future work, as it remains a distinct and standalone open challenge [27]. Moreover, for this figure we consider flat-fading channels in an analog MC system with M = 4 subcarriers.…”

Holographic MIMO Communications: Theoretical Foundations, Enabling Technologies, and Future Directions

“…It is noteworthy that, in far-field regions, Z tr can be neglected, i.e., Z tr ≈ 0, because the effects at the transmitting side caused by currents at the receiving side are negligible in far-field regions, forming the unilateral approximation relation [48]. However, this mutual impedance cannot be ignored if near-field regions are considered [244].…”

“…Exploiting the useful circuit theory based multi-port network, several recent works [244]- [249] were carried out with different emphases, such as communication models, near-field communications, channel modeling, and channel capacities.…”

Recent Advances in Machine Learning for Network Automation in the O-RAN

“…Finally, as latency remains one of the main challenges in Satellite NTN, reducing latency is important for Satellite NTN to support a wide range of applications. Some recent papers introduce the use of generative AI and digital twin in the communication system to deal with bandwidth limitation and, particularly, the latency [74]. As O-RAN offers convenient hosting of ML models, ML models can be pervasively deployed in O-RAN to encourage use of generative AI and digital twin in satellite communications.…”

“…Simulation tools such as OpenRAN Gym [83] and WiThRay [84] offer data generation for ML training when empirical data are insufficient or difficult to collect. However, data may be out-of-distribution due to different domains, and new learning techniques should be explored to tackle the out-of-distribution issue [85].…”