A Unified Convergence Analysis for Semi-Decentralized Learning: Sampled-to-Sampled vs. Sampled-to-All Communication

In semi-decentralized federated learning, devices primarily rely on device-to-device communication but occasionally interact with a central server. Periodically, a sampled subset of devices uploads their local models to the server, which computes an aggregate model. The server can then either (i) share this aggregate model only with the sampled clients (sampled-to-sampled, S2S) or (ii) broadcast it to all clients (sampled-to-all, S2A). Despite their practical significance, a rigorous theoretical and empirical comparison of these two strategies remains absent. We address this gap by analyzing S2S and S2A within a unified convergence framework that accounts for key system parameters: sampling rate, server aggregation frequency, and network connectivity. Our results, both analytical and experimental, reveal distinct regimes where one strategy outperforms the other, depending primarily on the degree of data heterogeneity across devices. These insights lead to concrete design guidelines for practical semi-decentralized FL deployments.

Joint Access Point Selection and Beamforming Design for Bistatic Backscatter Communication

Future Internet-of-Things networks are envisioned to use small and cheap sensor nodes with extremely low power consumption to avoid the extensive use of batteries. To provide connectivity to a massive number of these nodes, backscatter communication (BC) is emerging as an energy- and cost-efficient technology exploiting the reflection of radio frequency signals. However, challenges such as round-trip path loss and direct link interference (DLI) between the carrier emitter and the reader limit its performance. To tackle these limitations, this paper proposes a joint access point role selection and a novel beamforming technique for bistatic BC in a distributed multiple-input multiple-output setup. The proposed approach boosts the received backscattered energy while effectively mitigating DLI, thereby reducing the error probability. We also propose a channel estimation method tailored to operate under DLI conditions and propose a mismatch detector using estimated channel coefficients. Furthermore, we derive a closed-form expression for the probability of error for the detectors and model the quantization noise caused by DLI. Finally, comprehensive simulation results show that the proposed method with 1-bit analog-to-digital converters (ADCs) effectively mitigates DLI, reduces the quantization noise, and enhances backscattered signal energy, achieving performance comparable to the benchmark scenario with 8-bit ADCs.

DeMuon: A Decentralized Muon for Matrix Optimization over Graphs

In this paper, we propose DeMuon, a method for decentralized matrix optimization over a given communication topology. DeMuon incorporates matrix orthogonalization via Newton-Schulz iterations-a technique inherited from its centralized predecessor, Muon-and employs gradient tracking to mitigate heterogeneity among local functions. Under heavy-tailed noise conditions and additional mild assumptions, we establish the iteration complexity of DeMuon for reaching an approximate stochastic stationary point. This complexity result matches the best-known complexity bounds of centralized algorithms in terms of dependence on the target tolerance. To the best of our knowledge, DeMuon is the first direct extension of Muon to decentralized optimization over graphs with provable complexity guarantees. We conduct preliminary numerical experiments on decentralized transformer pretraining over graphs with varying degrees of connectivity. Our numerical results demonstrate a clear margin of improvement of DeMuon over other popular decentralized algorithms across different network topologies.

Industrial Viewpoints on RAN Technologies for 6G

6G standardization is to start imminently, with commercial deployments expected before 2030. Its technical components and performance requirements are the focus of this article. Our emphasis is on the 6G radio access, especially MIMO, AI, waveforms, coding, signal constellations and integration with non-terrestrial networks. Whilst standardization has not yet formally started, the scope of the 6G study items has been defined. Our predictions in this paper are speculative as there are no results of the study yet, but our views are guided by implementation and deployment aspects. We expect that the views here will guide researchers and industry practitioners.

Breaking the TDD Flow for Over-the-Air Phase Synchronization in Distributed Antenna Systems

Phase synchronization between distributed antenna arrays requires measurements that break the standard time-division duplex (TDD) operation. We present a feasibility study on implementing such synchronization and analyze its impact on the quality of service. Considering two antenna arrays with independent local oscillators (LOs), we propose a modified TDD flow to accommodate the transmission of phase synchronization signals, formulate the phase estimation and compensation problem, and derive the achievable downlink spectral efficiency (SE). Numerical results show that frequent re-estimation of the interarray phase disparity is essential for maximizing SE in systems with low-quality LOs. Furthermore, applying a Kalman filter for phase tracking substantially improves the SE, especially if phase estimation errors are large compared to LOs phase drifts.

Robust and Efficient Average Consensus with Non-Coherent Over-the-Air Aggregation

Non-coherent over-the-air (OTA) computation has garnered increasing attention for its advantages in facilitating information aggregation among distributed agents in resource-constrained networks without requiring precise channel estimation. A promising application scenario of this method is distributed average consensus in wireless multi-agent systems. However, in such scenario, non-coherent interference from concurrent OTA transmissions can introduce bias in the consensus value. To address this issue, we develop a robust distributed average consensus algorithm by formulating the consensus problem as a distributed optimization problem. Using decentralized projected gradient descent (D-PGD), our proposed algorithm can achieve unbiased mean square average consensus even in the presence of non-coherent interference and noise. Additionally, we implement transmit power control and receive scaling mechanisms to further accelerate convergence. Simulation results demonstrate that our method can significantly enhance the convergence speed of the D-PGD algorithm for OTA average consensus without compromising accuracy.

Impact of Network-Controlled Repeaters in Integrated Sensing and Communication Systems

Integrating sensing capabilities into existing massive MIMO communication networks has become crucial, stemming from a need for a more interconnected society. Improved coverage and performance can be obtained by incorporating new network components, such as reconfigurable intelligent surfaces or network-controlled repeaters (NCR). Integrating such components into modern networks brings a number of challenges. Thus, this paper contributes with the analysis of NCR impact in integrated sensing and communication networks. Particularly, the Cram\'er-Rao bound for a range estimator is derived where the interference from the repeater is taken into consideration. Additionally, a joint procedure for determining the repeater amplification factor, along with the precoders of the transmitting access point, is proposed.

Unified Analysis of Decentralized Gradient Descent: a Contraction Mapping Framework

The decentralized gradient descent (DGD) algorithm, and its sibling, diffusion, are workhorses in decentralized machine learning, distributed inference and estimation, and multi-agent coordination. We propose a novel, principled framework for the analysis of DGD and diffusion for strongly convex, smooth objectives, and arbitrary undirected topologies, using contraction mappings coupled with a result called the mean Hessian theorem (MHT). The use of these tools yields tight convergence bounds, both in the noise-free and noisy regimes. While these bounds are qualitatively similar to results found in the literature, our approach using contractions together with the MHT decouples the algorithm dynamics (how quickly the algorithm converges to its fixed point) from its asymptotic convergence properties (how far the fixed point is from the global optimum). This yields a simple, intuitive analysis that is accessible to a broader audience. Extensions are provided to multiple local gradient updates, time-varying step sizes, noisy gradients (stochastic DGD and diffusion), communication noise, and random topologies.

Physically Large Apertures for Wireless Power Transfer: Performance and Regulatory Aspects

Wireless power transfer (WPT) is a promising service for the Internet of Things, providing a cost-effective and sustainable solution to deploy so-called energy-neutral devices on a massive scale. The power received at the device side decays rapidly with the distance from a conventional transmit antenna with a physically small aperture. New opportunities arise from the transition from conventional far-field beamforming to near-field beam focusing. We argue that a "physically large" aperture, i.e., large w.r.t. the distance to the receiver, enables a power budget that remains practically independent of distance. Distance-dependent array gain patterns allow focusing the power density maximum precisely at the device location, while reducing the power density near the infrastructure. The physical aperture size is a key resource in enabling efficient yet regulatory-compliant WPT. We use real-world measurements to demonstrate that a regulatory-compliant system operating at sub-10GHz frequencies can increase the power received at the device into the milliwatt range. Our empirical demonstration shows that power-optimal near-field beam focusing inherently exploits multipath propagation, yielding both increased WPT efficiency and improved human exposure safety in real-world scenarios.

Whisper D-SGD: Correlated Noise Across Agents for Differentially Private Decentralized Learning

Decentralized learning enables distributed agents to train a shared machine learning model through local computation and peer-to-peer communication. Although each agent retains its dataset locally, the communication of local models can still expose private information to adversaries. To mitigate these threats, local differential privacy (LDP) injects independent noise per agent, but it suffers a larger utility gap than central differential privacy (CDP). We introduce Whisper D-SGD, a novel covariance-based approach that generates correlated privacy noise across agents, unifying several state-of-the-art methods as special cases. By leveraging network topology and mixing weights, Whisper D-SGD optimizes the noise covariance to achieve network-wide noise cancellation. Experimental results show that Whisper D-SGD cancels more noise than existing pairwise-correlation schemes, substantially narrowing the CDP-LDP gap and improving model performance under the same privacy guarantees.