Variational particle-based Bayesian learning methods have the advantage of not being limited by bias affecting more conventional parametric techniques. This paper proposes to leverage flexibility non-parametric approximate inference develop a novel federated unlearning method, referred as Forget-Stein Gradient Descent (Forget-SVGD). Forget-SVGD builds on SVGD – scheme using gradient-based deterministic updates and its distributed (federated) extension known Distributed (DSVGD). Upon...

In this Letter, the authors derive closed‐form formulas for generalised moment generating function of Fisher–Snedecor (F–S ) distribution, which enable us to effectively evaluate probability successful secure transmission over F–S composite fading channels. Extensive numerical results are presented validate accuracy their analytical framework.

Conventional frequentist federated learning (FL) schemes are known to yield overconfident decisions. Bayesian FL addresses this issue by allowing agents process and exchange uncertainty information encoded in distributions over the model parameters. However, comes at cost of a larger per-iteration communication overhead. This letter investigates whether can still provide advantages terms calibration when constraining bandwidth. We present compressed particle-based protocols for "unlearning"...

This correspondence considers the inverse Gaussian distribution, which is a tractable and accurate alternative to log-normal distribution that represents not only shadowing in wireless communications but also turbulence free-space optical communications. We first derive generalized moment generating function (G-MGF) of then make use derived formula effectively analyze several performance metrics, such as area under receiver operating characteristics (ROC) curve, average error rate, outage...

The authors formulate the closed-form expressions of generalised moment generating function (G-MGF) for extreme distribution, which enables one to calculate important metrics wireless communications systems. derived formula is utilised evaluate performance communication systems under fading channels, such as energy detection in terms area receiver operating characteristic curve and outage probability interference limited scenarios.

Federated learning enables edge devices to collaboratively train a global model while maintaining data privacy by keeping localized. However, the Non-IID nature of distribution across often hinders convergence and reduces performance. In this paper, we propose novel plugin for federated optimization techniques that approximates distributions IID through generative AI-enhanced augmentation balanced sampling strategy. Key idea is synthesize additional underrepresented classes on each device,...

Given sufficient data from multiple edge devices, federated learning (FL) enables training a shared model without transmitting private to central server. However, FL is generally vulnerable Byzantine attacks compromised which can significantly degrade the performance. In this paper, we propose intuitive plugin that be integrated into existing techniques achieve Byzantine-Resilience. Key idea generate virtual samples and evaluate consistency scores across local updates effectively filter out...

Federated Bayesian learning offers a principled framework for the definition of collaborative training algorithms that are able to quantify epistemic uncertainty and produce trustworthy decisions. Upon completion training, an agent may decide exercise her legal "right be forgotten", which calls contribution jointly trained model deleted discarded. This paper studies federated unlearning in decentralized network within framework. It specifically develops variational inference (VI) solutions...

Variational particle-based Bayesian learning methods have the advantage of not being limited by bias affecting more conventional parametric techniques. This paper proposes to leverage flexibility non-parametric approximate inference develop a novel federated unlearning method, referred as Forget-Stein Gradient Descent (Forget-SVGD). Forget-SVGD builds on SVGD - scheme using gradient-based deterministic updates and its distributed (federated) extension known Distributed (DSVGD). Upon...

This paper proposes a novel method to provide secrecy outage performance in various fading environments using mixture gamma (MG) distribution, which is well known approximate numerous channel models. First, we derive the generalized moment generating function (G-MGF) and incomplete G-MGF (IG-MGF) of MG distribution. Then, by applying those formulas, obtain exact asymptotic expressions probability along with corresponding lower bound strictly positive capacity for scenarios. Finally, accuracy...

In this letter, we present a new expression of ergodic capacity for two-wave with diffuse power (TWDP) fading channels. The derived formula is relatively concise and consists well-known functions even in infinite series form. Especially, the truncated approximate asymptotic are also presented, which enable us to obtain useful physical insights on effect TWDP various conditions.

Two-ray (TR) fading model is one of the models to represent a worst-case scenario. We derive exact closed-form expressions generalized moment generating function (G-MGF) for TR model, which enables us analyze numerous types wireless communication applications. Among them, we carry out several analytical results including ergodic capacity along with asymptotic and energy detection performance. Finally, provide numerical validate our evaluations.

Conventional frequentist FL schemes are known to yield overconfident decisions. Bayesian addresses this issue by allowing agents process and exchange uncertainty information encoded in distributions over the model parameters. However, comes at cost of a larger per-iteration communication overhead. This letter investigates whether can still provide advantages terms calibration when constraining bandwidth. We present compressed particle-based protocols for federated "unlearning" that apply...

This paper proposes a secure unmanned aerial vehicle (UAV) communication system with transmit-array antenna. We aim to maximize the average secrecy rate by jointly optimizing trajectory of UAV, phase shifters intelligent transmitting surface (ITS), and transmit beamforming vector. To address non-convexity formulated optimization problem, we apply an alternating (AO) technique. Simulation results verify proposed method demonstrate improved

Federated Bayesian learning offers a principled framework for the definition of collaborative training algorithms that are able to quantify epistemic uncertainty and produce trustworthy decisions. Upon completion training, an agent may decide exercise her legal "right be forgotten", which calls contribution jointly trained model deleted discarded. This paper studies federated unlearning in decentralized network within framework. It specifically develops variational inference (VI) solutions...