
In this paper, we study a bilinear saddle point problem of the form minx maxy F(x) + < Ax, y > - G(y), where F and G are & micro;F-and & micro;G-strongly convex functions, respectively. By incorporating Nesterov acceleration for strongly convex optimization, we first propose a first-order primal-dual gradient algorithm. We show that it achieves the linear convergence rate O((1-min{root & micro;F/LF, root & micro;G/LG})k) for both the primal-dual gap and the iterative gap, where LF and LG denote the smoothness constants of F and G, respectively. We further develop a continuous-time accelerated primal-dual dynamical system with constant damping. Using the Lyapunov analysis method, we establish the linear convergence rate O(e-min{root & micro;F,root & micro;G}t). Notably, when A = 0, our methods recover the classical Nesterov accelerated methods for strongly convex unconstrained problems in both discrete and continuous-time. Numerical experiments are presented to support the theoretical convergence results (c) 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
This study proposes a deployment-oriented reduced-order modeling framework for the rapid reconstruction of aerated gas-liquid two-phase flow fields in a pump-turbine. The framework integrates high-fidelity computational fluid dynamics (CFD), proper orthogonal decomposition (POD), a feedforward neural network (FNN), and ANSYS Twin Builder to enable rapid prediction of pressure and velocity fields under coupled inlet liquid massflow-rate and aeration conditions. CFD results obtained from different combinations of inlet liquid mass flow rate and aeration velocity were used to construct separate snapshot matrices for the pressure and velocity fields. POD was then applied to extract the dominant flow-field modes, and FNN surrogate models were trained to establish the nonlinear mapping between operating parameters and the corresponding modal coefficients. The reconstructed results show that the proposed POD-FNN model can reproduce the main spatial distribution characteristics of the pressure and velocity fields in key regions, including the guide-vane passages and the runner. For the independent validation samples, the mean and maximum relative errors of head prediction were 0.172% and 0.277%, respectively. After deployment in ANSYS Twin Builder, the response time for selected monitoring-point outputs was approximately 40 ms, and the reconstruction time for a pressure or velocity field was about 1 s, Whereas a single high-fidelity CFD simulation required approximately 12 h. These results indicate that the proposed reduced-order model can serve as a rapid-inference computational kernel for future digital-twin applications of pump-turbine flow-field prediction.
Federated Learning (FL) enables collaborative model training without centralizing raw data and has become an important paradigm for privacy-sensitive intelligent systems. Nevertheless, its distributed architecture, heterogeneous data, and unreliable participants introduce substantial robustness and security challenges, ranging from communication noise and system failures to poisoning, privacy inference, and Byzantine attacks. In this paper, we present a structured survey of robust federated learning. We first organize the literature through a threat-oriented taxonomy that covers poisoning attacks, privacy leakage risks, Byzantine behavior, and scenario-specific robustness challenges. We explicitly analyze stronger adversaries that collude, coordinate poisoning campaigns, replace models behind secure aggregation, and adapt their optimization to defense-specific weaknesses. We then review representative defense mechanisms, including robust aggregation, backdoor defense, privacy-preserving training, secure aggregation, and personalized robustness enhancement. Particular attention is given to three recurring tensions in the literature: robustness under Non-IID data, robustness-efficiency trade-offs, and robustness-privacy trade-offs. We further discuss robust federated learning in wireless and IoT environments, where communication uncertainty and resource constraints become first-order concerns. We examine how existing robustness mechanisms translate, or fail to translate, to foundation-model-based FL, federated fine-tuning of large language models, federated reinforcement learning, agentic federated systems, and federated retrieval-augmented generation. Finally, we summarize open challenges and outline future research directions for building more reliable, efficient, and deployable federated learning systems.
Drone delivery, as a low-altitude logistics application of urban air mobility (UAM), may reshape last-mile distribution and urban service accessibility. However, its societal diffusion cannot be fully understood through conventional binary acceptance models, because consumers may differ substantially in their readiness to progress from awareness to repeated use. This study investigates heterogeneous consumer readiness for drone delivery using 1,018 valid survey responses collected in Shenzhen, China, where urban drone delivery has entered early deployment. Drawing on the Transtheoretical Model, adoption is conceptualized as an ordered five-stage readiness process: precontemplation, contemplation, preparation, action, and maintenance. An integrated framework combining the Technology Acceptance Model and the Theory of Planned Behavior, extended by environmental concern and perceived risk, is used to identify attitudinal latent constructs. A hybrid model integrating a Multiple Indicators and Multiple Causes model and a Partial Proportional Odds Model examines how socio-demographic attributes and psychological perceptions shape stage-specific readiness, with marginal effects quantifying heterogeneous influences across stages. Results show that drone delivery adoption is not a linear extension of general acceptance but a staged readiness process shaped by different psychological mechanisms. Socio-demographic attributes mainly explain baseline differences in adoption propensity, while attitudinal constructs exert stronger and more stage-dependent effects. Early-stage readiness is constrained by perceived risk and uncertainty, but is also sensitive to social influence and environmental concern; progression toward action and maintenance is more closely associated with perceived usefulness, ease of use, perceived behavioral control, attitude, and behavioral intention. This pattern indicates a pathway from risk reduction and cognitive activation to technology evaluation, experiential confidence, and behavioral consolidation. Findings provide behavioral evidence for stage-targeted risk communication, service design, and demand cultivation strategies in early urban drone delivery deployment.
Reconstructing structural dynamic responses from limited measurements under unknown excitations is of great importance in structural dynamics. Finite element modeling and in-situ response measurements at selected locations are normally required, and therefore measurement noise and finite element model errors can jointly affect reconstruction accuracy and confidence. Reconstruction accuracy is also highly dependent on sensor placement. To address these issues, this paper proposes a time‑domain response reconstruction method based on Bayesian inference that simultaneously accounts for both uncertainties and optimizes sensor placement to minimize their effects. First, based on the modal superposition theory and the Bayesian framework, a joint prior distribution of modal coordinates and mode shape error vectors is constructed, and its posterior distribution is then derived, thereby achieving optimal estimation and uncertainty quantification of the response reconstruction. On this basis, the modal kinetic energy criterion and a clustering algorithm are incorporated to establish a sensor placement optimization method that maximizes reconstruction accuracy and minimizes information redundancy, thereby reducing reconstruction uncertainty. Subsequently, numerical examples are used to analyze the reconstruction accuracy under typical excitations, and the effects of sensor number, measurement noise, and model errors are discussed. Comparisons with representative classical sensor-placement methods are also conducted to assess the proposed strategy. Finally, experiments on an aluminum profile under transient excitation are conducted to validate the proposed method.