Quantum Sparse Autoencoders for Q-Matrix Estimation in Cognitive Diagnosis
本文针对认知诊断中Q矩阵估计难题，提出了一种新的量子稀疏自动编码器(QSAE)方法，并通过与经典自动编码器对比验证了其在处理真实数据集时的稳定性和探索潜在结构复杂性的能力。
0 citationsRead paper
Selected work
Representative Papers
A Machine Learning Based Search for Lunar Anomalies
This study addresses the challenge of automatically detecting scientific or anthropogenic anomalous features in vast collections of high-resolution lunar imagery. It proposes an unsupervised anomaly detection method by introducing Beta-variational autoencoders (Beta-VAE) to large-scale lunar remote sensing data—a first in this domain. The approach requires no labeled training data and simultaneously identifies both natural geological anomalies and artificial objects. Applied to Lunar Reconnaissance Orbiter (LRO) images acquired since 2009, the model successfully locates scientifically significant craters such as Plaskett and Paracelsus C and accurately pinpoints multiple known lander sites with statistical significance. These results demonstrate the method’s effectiveness and generalization capability for anomaly detection in planetary remote sensing.
0 citationsRead paper
ASTELD: A Six-Axis Classification Framework for Autonomous AI Agents - Design, Evaluation, and an OpenClaw Case Study
Current autonomous AI agent platforms exhibit significant heterogeneity in architecture, security, and tool integration, yet lack a unified taxonomy to enable systematic comparison and design analysis. This work proposes ASTELD, a six-dimensional classification framework that establishes, for the first time, a rule-based multi-axis taxonomy encompassing architectural patterns, security postures, tool integration models, execution paradigms, levels of autonomy, and deployment topologies. By synthesizing existing taxonomies, analyzing platform attributes, mapping multiple platforms, and conducting a case study on OpenClaw, the framework successfully differentiates eight major platforms, identifies three cross-platform design patterns, classifies over fifty derivative systems, and reveals a critical design gap—namely, the absence of platforms combining local-first deployment with enterprise-grade security—thereby providing a reproducible methodological foundation for future agent platform comparison and research.
0 citationsRead paper
Optimistic Imprecise Shortest Watchtower in 1.5D and 2.5D
This study addresses the problem of selecting vertex heights within prescribed vertical intervals on uncertain 1.5D and 2.5D terrains to minimize the height of the shortest watchtower that can cover the entire terrain. The work presents the first linear-time exact algorithm for the 1.5D case and introduces an approximation scheme for the discrete 2.5D setting with additive error ε, running in O((OPT/ε)·n³) time. By integrating techniques for handling interval constraints, visibility analysis, and approximation algorithm design, this research overcomes key computational bottlenecks in visibility optimization under terrain uncertainty, substantially advancing the theoretical and algorithmic foundations of the problem.
0 citationsRead paper
Recent publications
Latest Papers
Quantum Sparse Autoencoders for Q-Matrix Estimation in Cognitive Diagnosis
本文针对认知诊断中Q矩阵估计难题，提出了一种新的量子稀疏自动编码器(QSAE)方法，并通过与经典自动编码器对比验证了其在处理真实数据集时的稳定性和探索潜在结构复杂性的能力。
0 citationsRead paper
A Machine Learning Based Search for Lunar Anomalies
This study addresses the challenge of automatically detecting scientific or anthropogenic anomalous features in vast collections of high-resolution lunar imagery. It proposes an unsupervised anomaly detection method by introducing Beta-variational autoencoders (Beta-VAE) to large-scale lunar remote sensing data—a first in this domain. The approach requires no labeled training data and simultaneously identifies both natural geological anomalies and artificial objects. Applied to Lunar Reconnaissance Orbiter (LRO) images acquired since 2009, the model successfully locates scientifically significant craters such as Plaskett and Paracelsus C and accurately pinpoints multiple known lander sites with statistical significance. These results demonstrate the method’s effectiveness and generalization capability for anomaly detection in planetary remote sensing.
0 citationsRead paper
ASTELD: A Six-Axis Classification Framework for Autonomous AI Agents - Design, Evaluation, and an OpenClaw Case Study
Current autonomous AI agent platforms exhibit significant heterogeneity in architecture, security, and tool integration, yet lack a unified taxonomy to enable systematic comparison and design analysis. This work proposes ASTELD, a six-dimensional classification framework that establishes, for the first time, a rule-based multi-axis taxonomy encompassing architectural patterns, security postures, tool integration models, execution paradigms, levels of autonomy, and deployment topologies. By synthesizing existing taxonomies, analyzing platform attributes, mapping multiple platforms, and conducting a case study on OpenClaw, the framework successfully differentiates eight major platforms, identifies three cross-platform design patterns, classifies over fifty derivative systems, and reveals a critical design gap—namely, the absence of platforms combining local-first deployment with enterprise-grade security—thereby providing a reproducible methodological foundation for future agent platform comparison and research.
0 citationsRead paper
Optimistic Imprecise Shortest Watchtower in 1.5D and 2.5D
This study addresses the problem of selecting vertex heights within prescribed vertical intervals on uncertain 1.5D and 2.5D terrains to minimize the height of the shortest watchtower that can cover the entire terrain. The work presents the first linear-time exact algorithm for the 1.5D case and introduces an approximation scheme for the discrete 2.5D setting with additive error ε, running in O((OPT/ε)·n³) time. By integrating techniques for handling interval constraints, visibility analysis, and approximation algorithm design, this research overcomes key computational bottlenecks in visibility optimization under terrain uncertainty, substantially advancing the theoretical and algorithmic foundations of the problem.
0 citationsRead paper