Hybrid Variational Quantum Circuits for Multivariate Regression and High-Dimensional Data Reconstruction
本文提出了一种混合变分量子电路(HVQC)，通过结合经典仿射后测量层，实现多变量回归和高维数据重建，实验效果优于传统方法。
0 citationsRead paper
Selected work
Representative Papers
Path Tracking with Dynamic Control Point Blending for Autonomous Vehicles: An Experimental Study
This work addresses the limited path-tracking accuracy and stability in autonomous driving under low-speed and reverse maneuvers caused by fixed control points (front or rear axle). To overcome this, the authors propose a dynamic control-point fusion framework that continuously interpolates between the outputs of a front-axle Stanley controller and a rear-axle curvature-based geometric controller based on the vehicle’s center of gravity. This lateral control strategy is synergistically integrated with a longitudinal velocity modulation scheme that leverages curvature-aware virtual track boundaries and ray-casting techniques. The key innovation lies in the continuous interpolation mechanism, enabling smooth transitions between front- and rear-axle control modes. Experimental results from both simulation and real-world vehicle tests demonstrate that the proposed approach significantly improves trajectory tracking accuracy, steering smoothness, and adaptability in complex maneuvers such as closed-loop tracking and reversing, outperforming conventional fixed-control-point methods.
0 citationsRead paper
On the Approximation Power of SiLU Networks: Exponential Rates and Depth Efficiency
This work investigates the approximation power of SiLU-activated neural networks for smooth functions, aiming to overcome the polynomial convergence rate limitation inherent in ReLU networks. We propose a hierarchical construction based on efficient quadratic approximations, achieving— for the first time with constant depth (O(1))—an approximation error decay of O(ω⁻²ᵏ). We establish tight depth–size trade-offs for approximating Sobolev functions, attaining parameter complexity O(ε⁻ᵈ⁄ⁿ), which significantly improves upon comparable ReLU-based constructions. Theoretically, we prove that SiLU networks achieve exponential approximation rates while maintaining constant depth and optimal size scaling. This is the first systematic theoretical characterization revealing SiLU’s intrinsic advantage in approximating smooth functions, providing rigorous justification for activation function selection in deep learning.
0 citationsRead paper
AUTOPSY: A Framework for Tackling Privacy Challenges in the Automotive Industry
Intelligent connected vehicles (ICVs) face urgent practical challenges in enhancing privacy-friendliness beyond mere GDPR compliance. Method: This study proposes the first privacy engineering framework tailored for full-vehicle systems, integrating system modeling, a dynamic privacy manager, a GDPR principle–guided PETs selection methodology, and a layered privacy architecture to enable fine-grained data-flow control and automated compliance mapping. The modular design decouples privacy functionality for scalable deployment. A prototype is implemented and validated in a location-based service scenario. Contribution/Results: The framework ensures end-to-end controllability over user data collection, transmission, and processing; achieves privacy-policy response latency under 200 ms; and improves PETs configuration coverage by 40%. It establishes a technically advanced, regulation-adaptive privacy enhancement paradigm for automotive systems.
0 citationsRead paper
Unlocking the Potential of Generative AI through Neuro-Symbolic Architectures: Benefits and Limitations
This study addresses key bottlenecks in generative AI—limited generalization, weak reasoning capability, poor interpretability, and low data efficiency. To this end, we propose a novel bidirectional coupled neuro-symbolic architecture (Neuro>Symbolic<Neuro), which seamlessly integrates deep learning with symbolic reasoning. Methodologically, we design a unified collaborative framework that synergistically incorporates retrieval-augmented generation, graph neural networks, reinforcement learning, and multi-agent systems, enabling dynamic, bidirectional interaction and mutual enhancement between neural and symbolic modules. Experimental results demonstrate substantial improvements in generalization, structured reasoning, interpretability, and data efficiency on complex tasks such as logical reasoning and trustworthy content generation. The architecture achieves state-of-the-art performance across multiple benchmarks, establishing a new paradigm for trustworthy, efficient, and scalable generative AI.
0 citationsRead paper
Recent publications
Latest Papers
Hybrid Variational Quantum Circuits for Multivariate Regression and High-Dimensional Data Reconstruction
本文提出了一种混合变分量子电路(HVQC)，通过结合经典仿射后测量层，实现多变量回归和高维数据重建，实验效果优于传统方法。
0 citationsRead paper
Path Tracking with Dynamic Control Point Blending for Autonomous Vehicles: An Experimental Study
This work addresses the limited path-tracking accuracy and stability in autonomous driving under low-speed and reverse maneuvers caused by fixed control points (front or rear axle). To overcome this, the authors propose a dynamic control-point fusion framework that continuously interpolates between the outputs of a front-axle Stanley controller and a rear-axle curvature-based geometric controller based on the vehicle’s center of gravity. This lateral control strategy is synergistically integrated with a longitudinal velocity modulation scheme that leverages curvature-aware virtual track boundaries and ray-casting techniques. The key innovation lies in the continuous interpolation mechanism, enabling smooth transitions between front- and rear-axle control modes. Experimental results from both simulation and real-world vehicle tests demonstrate that the proposed approach significantly improves trajectory tracking accuracy, steering smoothness, and adaptability in complex maneuvers such as closed-loop tracking and reversing, outperforming conventional fixed-control-point methods.
0 citationsRead paper
On the Approximation Power of SiLU Networks: Exponential Rates and Depth Efficiency
This work investigates the approximation power of SiLU-activated neural networks for smooth functions, aiming to overcome the polynomial convergence rate limitation inherent in ReLU networks. We propose a hierarchical construction based on efficient quadratic approximations, achieving— for the first time with constant depth (O(1))—an approximation error decay of O(ω⁻²ᵏ). We establish tight depth–size trade-offs for approximating Sobolev functions, attaining parameter complexity O(ε⁻ᵈ⁄ⁿ), which significantly improves upon comparable ReLU-based constructions. Theoretically, we prove that SiLU networks achieve exponential approximation rates while maintaining constant depth and optimal size scaling. This is the first systematic theoretical characterization revealing SiLU’s intrinsic advantage in approximating smooth functions, providing rigorous justification for activation function selection in deep learning.
0 citationsRead paper
AUTOPSY: A Framework for Tackling Privacy Challenges in the Automotive Industry
Intelligent connected vehicles (ICVs) face urgent practical challenges in enhancing privacy-friendliness beyond mere GDPR compliance. Method: This study proposes the first privacy engineering framework tailored for full-vehicle systems, integrating system modeling, a dynamic privacy manager, a GDPR principle–guided PETs selection methodology, and a layered privacy architecture to enable fine-grained data-flow control and automated compliance mapping. The modular design decouples privacy functionality for scalable deployment. A prototype is implemented and validated in a location-based service scenario. Contribution/Results: The framework ensures end-to-end controllability over user data collection, transmission, and processing; achieves privacy-policy response latency under 200 ms; and improves PETs configuration coverage by 40%. It establishes a technically advanced, regulation-adaptive privacy enhancement paradigm for automotive systems.
0 citationsRead paper
Unlocking the Potential of Generative AI through Neuro-Symbolic Architectures: Benefits and Limitations
This study addresses key bottlenecks in generative AI—limited generalization, weak reasoning capability, poor interpretability, and low data efficiency. To this end, we propose a novel bidirectional coupled neuro-symbolic architecture (Neuro>Symbolic<Neuro), which seamlessly integrates deep learning with symbolic reasoning. Methodologically, we design a unified collaborative framework that synergistically incorporates retrieval-augmented generation, graph neural networks, reinforcement learning, and multi-agent systems, enabling dynamic, bidirectional interaction and mutual enhancement between neural and symbolic modules. Experimental results demonstrate substantial improvements in generalization, structured reasoning, interpretability, and data efficiency on complex tasks such as logical reasoning and trustworthy content generation. The architecture achieves state-of-the-art performance across multiple benchmarks, establishing a new paradigm for trustworthy, efficient, and scalable generative AI.
0 citationsRead paper