High-dimensional studies often contain a small set of prominent signals embedded in a broad non-sparse background, which can undermine purely sparse modeling. This paper proposes Weighted Non-sparse and Sparse Iteration (WNSI), an iterative procedure for joint estimation of sparse and non-sparse components that incorporates adaptive reweighting to stabilize updates and improve adaptivity across iterations. WNSI alternates between precision matrix-based estimation for the non-sparse component and an adaptively weighted penalized regression step for the sparse component, balancing numerical stability and data-driven refinement. We prove that WNSI converges to the oracle solution and establish non-asymptotic error bounds, achieving optimal l2 rates under standard regularity conditions. Simulation studies show that WNSI achieves improved estimation accuracy and variable selection, and remains robust under strong dependence and distributional misspecification. An application to breast cancer gene expression data illustrates that WNSI achieves competitive prediction across multiple screening and grouping configurations, highlighting its practical utility for high-dimensional genomic systems.
Nudge refers to the lateral detouring behavior of autonomous vehicles around static traffic participants. In autonomous robo-delivery systems, nudge decision-making is a frequent and critical classification problem, where suboptimal decisions lead to safety hazards and delivery delays. The inherent characteristics of nudge decision-making process, including high-dimensional spatiotemporal features and a continuous state-action space, make it particularly challenging for traditional rule-based approaches. These methods struggle to devise exhaustive rules and require ongoing optimization for long-tail scenarios. Data-driven methods, in contrast, excel at learning decision boundaries in such complex decision spaces and demonstrate superior generalization capabilities. To leverage these advantages, we developed a transformer architecture for high-dimensional, continuous state-action space decision classification problems, comprising a general-purpose encoder for extracting comprehensive spatiotemporal features, and a multi-head decoder that enables joint learning of multi-obstacle decision-making and multi-modal trajectory generation specifically tailored for nudge tasks. To ensure robust generalization, we curated a large-scale, high-quality nudge-specific dataset with 128 million scene samples, obtained through data mining, distillation, and auto-labeling, and exhibiting a uniform distribution across both temporal and spatial dimensions. Experimental results demonstrate that our model achieves a decision average precision of 99.9%, a trajectory average displacement error of 1.729 m, and an inference latency of 3.1 ms. It shows a 13.55% relative improvement in simulation-based evaluation pass rate compared to the rule-based method, while also outperforming representative learning-based baselines on a common dataset and evaluation split, with further validation through 5 million kilometers of road testing.
Cloud-based Machine Learning as a Service (CMLaaS) platforms provide businesses with on-demand machine learning capabilities, accelerating the commercial adoption of ML technologies. However, these platforms are vulnerable to poisoning attacks that compromise service integrity. In this paper, we introduce a service-attack framework based on a retrial queueing model to evaluate CMLaaS performance in the presence of such attacks. In our model, external adversaries launch targeted poisoning attempts that remain undetected unless specifically checked; if an attack is not neutralized (i.e., repaired) before service completion, the resulting output is deemed invalid. To address the need for rapid and precise reliability assessment in engineering practice, this paper introduces an algorithm based on Monte Carlo simulation to estimate instantaneous reliability. Then, the strategic behavior of customers in self-interested and altruistic cases is analyzed, revealing the important role of curbing selfish behavior in improving system reliability. In addition, the optimization problem of the cost-effectiveness ratio (CER) under different cost structures is studied to provide a more comprehensive analysis of operational strategies for CMLaaS platforms with different purposes. Finally, the benchmark model is extended to the cases of imperfect detection, imperfect repair and finite capacity. The model is validated through numerical experiments, demonstrating the necessity of enhancing repair capability.
China’s low-carbon transition requires the effective allocation of green innovation factors, yet their spatial imbalance constrains national innovation efficiency. This study examines how the agglomeration of green finance reshapes the flow of green innovation factors by distinguishing between the level and speed of agglomeration. The results show that moderate agglomeration of green finance promotes the interregional circulation of green innovation factors, while excessively rapid agglomeration hinders their spillover effects. Furthermore, the development of technology markets and regional economic conditions are found to exert significant moderating influences, revealing a nonlinear relationship between green finance agglomeration and green innovation factor flow. These insights enrich the understanding of interactions between finance and innovation during economic restructuring, and offer policy insights for promoting balanced green development.
The existing literature has not reached a consensus on the relationship and mechanisms between innovation and income inequality, particularly in the context of urban–rural disparities. To address this issue, we investigate the conditions under which innovation increases or decreases income inequality. We propose that the impact of innovation on income inequality varies across different stages of innovation development. Using provincial-level panel data from China spanning 2009 to 2020, we constructed multiple indicators to measure innovation and found that the impact of innovation on urban–rural income inequality exhibits an inverted U-shaped trend: innovation initially exacerbates urban–rural income inequality and subsequently alleviates it. This phenomenon is influenced by mechanisms such as the learning-by-doing effect, the erosion effect, and industrial structure upgrading. Additionally, our research shows that regional differences in the intensity of these mechanisms account for the heterogeneity in the impact of innovation on income inequality. This study contributes to a deeper understanding of the dual effects of innovation on socio-economic disparities.
