Federated learning (FL) has emerged as a promising distributed learning framework that preserves data privacy on edge devices. Over-the-air federated learning (OTA-FL) considers the shared multiple access medium and leverages the superposition property of wireless channels to aggregate local updates, making it particularly well-suited for communication-efficient FL. In this paper, we consider energy harvesting edge devices as clients and develop participation (client-selection) mechanisms for energy efficient OTA-FL in a distributed manner, i.e., each edge device decides whether to participate in the global learning iteration based on its local knowledge. Specifically, each edge device independently decides on participation in a learning round under imperfect CSI, non-i.i.d. data, and fading channels. By defining a device importance metric that captures the contribution of each device to global loss and gradient divergence, we formulate a Lyapunov online optimization framework, allowing each device to balance local computation steps with harvested energy constraints. We provide the convergence analysis for the proposed algorithm. Numerical results demonstrate that our solution significantly outperforms baselines, highlighting its effectiveness for large-scale, heterogeneous, and energy-constrained FL deployments.

Over-the-air federated learning (OTA-FL) provides bandwidth-efficient learning by leveraging the inherent superposition property of wireless channels. Personalized federated learning balances performance for users with diverse datasets, addressing real-life data heterogeneity. We propose the first personalized OTA-FL scheme through multi-task learning, assisted by personal reconfigurable intelligent surfaces (RIS) for each user. We take a cross-layer approach that optimizes communication and computation resources for global and personalized tasks in time-varying channels with imperfect channel state information, using multi-task learning for non-i.i.d data. Our PROAR-PFed algorithm adaptively designs power, local iterations, and RIS configurations. We present convergence analysis for non-convex objectives and demonstrate that PROAR-PFed outperforms state-of-the-art on the Fashion-MNIST dataset.

Federated learning (FL) is a distributed learning framework that by design allows local edge devices to keep their training data. However, privacy leakage occurs through model updates and is a privacy protection concern that needs to be addressed. Over-the-air FL (OTA-FL) is a variant of FL designed for wireless edge networks by utilizing the inherent superposition property of the wireless medium. The wireless physical layer (PHY), in addition to providing resource and communication-efficient collaborative training via OTA-FL, can also be leveraged to enhance privacy for FL. This paper presents the PHY design to ensure differentially private (DP) OTA-FL. Specifically, by leveraging the Gaussian noise naturally present in the wireless channel, and deploying a dedicated artificial noise generator (cooperative jammer) when needed, a fully decentralized, dynamic power control strategy is proposed. This design relies on a resource-efficient FL framework with first-order approximation applied at every even iteration, thereby reducing the amount of information needed from clients. This approach can eliminate the need for artificial noise injection at the client side, typically required to achieve DP; the cooperative jammer is used for higher privacy requirement without transmission efficiency loss. The privacy analysis is provided via the Moments Accountant method, providing a tight privacy assessment. The convergence analysis is provided for non-convex learning objectives. Experiments conducted on real-world non-i.i.d. data demonstrate that our scheme outperforms the state-of-the-art method under the same DP requirement and illustrate the effectiveness of cooperative jammer in the case of stringent privacy requirements.

Over-the-air federated learning (OTA-FL) integrates communication and model aggregation by exploiting the innate superposition property of wireless channels. The approach renders bandwidth efficient learning, but requires care in handling the wireless physical layer impairments. In this paper, federated edge learning is considered for a network that is heterogeneous with respect to client (edge node) data set distributions and individual client resources, under a general non-convex learning objective. We augment the wireless OTA-FL system with a Reconfigurable Intelligent Surface (RIS) to enable a propagation environment with improved learning performance in a realistic time varying physical layer. Our approach is a cross-layer perspective that jointly optimizes communication, computation and learning resources, in this general heterogeneous setting. We adapt the local computation steps and transmission power of the clients in conjunction with the RIS phase shifts. The resulting joint communication and learning algorithm, RIS-assisted Over-the-air Adaptive Resource Allocation for Federated learning (ROAR-Fed) is shown to be convergent in this general setting. Numerical results demonstrate the effectiveness of ROAR-Fed under heterogeneous (non i.i.d.) data and imperfect CSI, indicating the advantage of RIS assisted learning in this general set up.

Reconfigurable Intelligent Surfaces (RIS) are a new paradigm which, with judicious deployment and alignment, can enable more favorable propagation environments and better wireless network design. As such, they can offer a number of potential benefits for next generation wireless systems including improved coverage, better interference management and even security. In this paper, we consider an uplink next generation wireless system where each user is assisted with an RIS. We study the uplink power control problem in this distributed RIS-assisted wireless network. Specifically, we aim to minimize total uplink transmit power of all the users subject to each user's reliable communication requirements at the base station by a joint design of power, receiver filter and RIS phase matrices. We propose an iterative power control algorithm, combined with a successive convex approximation technique to solve the problem with non-convex phase constraints. Numerical results illustrate that distributed RIS assistance leads to uplink power savings when direct links are weak.

Rapid Serial Visual Presentation (RSVP) is an important paradigm in Brain-Computer Interface (BCI). It can be used in speller, image retrieval, anomaly detection, etc. RSVP paradigm uses a small number of target pictures in a high speed presented picture sequence to induce specific event-related potential (ERP) components. However, the application of RSVP based BCI is challenged by the accuracy of ERP detection. Thus, the goal of this study is to introduce other related modalities to the traditional EEG-based BCI to make robust predictions and improve the detection performance. First, we introduce the eye movement modality into the RSVP-based BCI and collect a multimodality RSVP-based dataset simultaneously during the image retrieval task. Second, we design a simple but efficient CNN-based network with two modality fusion modules to fully utilize the multi-modality data in two stages. In the feature extraction stage, we propose a Cross-modality-Guided Feature Calibration (cm-GFC) module to enable the EEG modality feature to modify the eye movement modality feature, and the aim is to make eye movement modality features and EEG modality features are more complementary. In the feature fusion stage, we propose a Dynamic Gated Fusion (DGF) module, which applies modality-specific gates to retain the complementary information of the two modalities and reduce redundant information from the two modalities. To evaluate our method, we conduct extensive experiments on the dataset with EEG and eye movement data are from 20 subjects. The proposed method achieves a high balanced accuracy of 87.83 ± 2.31% of classification, which outperforms a series of single modality and multi-modality approaches.