Parameter Estimation for the Mixed Fractional Merton Jump Diffusion Model with EM Algorithm
本文针对混合分数Merton跳跃扩散模型，提出了一种结合Metropolis-Hastings采样的EM算法进行参数估计，以捕捉金融回报中的长程依赖性和跳跃行为。
0 citationsRead paper
Selected work
Representative Papers
Pricing Temperature-Index Insurance under Long Memory and Stochastic Time Change
This study addresses the challenge of neglecting long memory and stochastic time-varying characteristics in temperature index insurance pricing by constructing an actuarial framework based on fractional Brownian motion and the CIR process. By deriving conditional Gaussian representations and exact kernel functions, combined with Monte Carlo simulation, the proposed approach achieves efficient valuation while avoiding complex path simulations. Empirical results demonstrate that this model significantly enhances the accuracy of climate risk pricing and reveals the substantial impact of long memory and time-varying properties on premiums. Consequently, this work provides innovative methodological support for advancing weather derivative pricing theory, effectively bridging the gap between sophisticated stochastic modeling and practical actuarial applications in managing temperature-related financial risks.
0 citationsRead paper
Climate-Conditioned Cascade Modeling for Multi-Peril Reinsurance: Analysis and Controlled Numerical Applications
Traditional joint loss models struggle to capture the temporal dependencies and state-dependent propagation among climate hazards, leading to biased reinsurance risk assessments. This work proposes a Cascading Climate Risk Network (CCRN) that decouples annual-scale climatic conditions from intra-event propagation mechanisms using a directed acyclic graph. The model maps physical states to insurance losses through a complementary log-log triggering function, bounded severity activation, and a capacity-constrained demand surge transformation. It innovatively derives a path-dependent closed-form solution for cascading losses and constructs pathwise upper-bound losses over rectangular stress sets, enabling transparent, contract-level stress testing. Numerical experiments, conducted for the first time in a synthetic environment, confirm that directional propagation critically shapes tail risk, identifying directional propagation, annual event frequency, and dependency strength as the three key drivers. Mid-layer reinsurance pricing proves robust to marginal dependency structures, whereas upper-tail behaviors exhibit significant divergence.
0 citationsRead paper
Generative AI for Quantum Circuits and Quantum Code: A Technical Review and Taxonomy
This study addresses the lack of a unified evaluation framework for generative AI in quantum circuit and code generation, particularly the absence of validation on real quantum hardware. Through a structured scoping review, the authors systematically analyze 13 generative systems and five datasets, proposing the first taxonomy based on output modality and training paradigm. They further introduce a three-tiered evaluation framework encompassing syntactic validity, semantic correctness, and hardware executability. The findings reveal that while existing methods generally satisfy syntactic requirements and partially meet semantic criteria, none demonstrate end-to-end validation on actual quantum devices. This critical gap underscores the field’s current limitation in real-device verification and provides a clear direction for future research toward practical, hardware-aware quantum program synthesis.
0 citationsRead paper
Malware Detection based on API Calls: A Reproducibility Study
This study validates and reproduces a malware detection approach based on API call frequency, with a focus on its reproducibility and robustness. Using a Random Forest classifier, the authors construct four model variants—Unigram, Bigram, Trigram, and combined n-grams—and evaluate them on API sequences of length 2,500. The experiments independently confirm for the first time that the original method exhibits high performance stability across different random seeds, with a standard deviation below 0.5% over three trials. Moreover, all model variants achieve F1 scores surpassing those reported in the original study by 0.99% to 2.57%, with the Unigram model reaching an F1 score of 0.8717. These results significantly enhance the scientific rigor and practical viability of the API-based malware detection methodology.
0 citationsRead paper
Recent publications
Latest Papers
Parameter Estimation for the Mixed Fractional Merton Jump Diffusion Model with EM Algorithm
本文针对混合分数Merton跳跃扩散模型，提出了一种结合Metropolis-Hastings采样的EM算法进行参数估计，以捕捉金融回报中的长程依赖性和跳跃行为。
0 citationsRead paper
Pricing Temperature-Index Insurance under Long Memory and Stochastic Time Change
This study addresses the challenge of neglecting long memory and stochastic time-varying characteristics in temperature index insurance pricing by constructing an actuarial framework based on fractional Brownian motion and the CIR process. By deriving conditional Gaussian representations and exact kernel functions, combined with Monte Carlo simulation, the proposed approach achieves efficient valuation while avoiding complex path simulations. Empirical results demonstrate that this model significantly enhances the accuracy of climate risk pricing and reveals the substantial impact of long memory and time-varying properties on premiums. Consequently, this work provides innovative methodological support for advancing weather derivative pricing theory, effectively bridging the gap between sophisticated stochastic modeling and practical actuarial applications in managing temperature-related financial risks.
0 citationsRead paper
Climate-Conditioned Cascade Modeling for Multi-Peril Reinsurance: Analysis and Controlled Numerical Applications
Traditional joint loss models struggle to capture the temporal dependencies and state-dependent propagation among climate hazards, leading to biased reinsurance risk assessments. This work proposes a Cascading Climate Risk Network (CCRN) that decouples annual-scale climatic conditions from intra-event propagation mechanisms using a directed acyclic graph. The model maps physical states to insurance losses through a complementary log-log triggering function, bounded severity activation, and a capacity-constrained demand surge transformation. It innovatively derives a path-dependent closed-form solution for cascading losses and constructs pathwise upper-bound losses over rectangular stress sets, enabling transparent, contract-level stress testing. Numerical experiments, conducted for the first time in a synthetic environment, confirm that directional propagation critically shapes tail risk, identifying directional propagation, annual event frequency, and dependency strength as the three key drivers. Mid-layer reinsurance pricing proves robust to marginal dependency structures, whereas upper-tail behaviors exhibit significant divergence.
0 citationsRead paper
Generative AI for Quantum Circuits and Quantum Code: A Technical Review and Taxonomy
This study addresses the lack of a unified evaluation framework for generative AI in quantum circuit and code generation, particularly the absence of validation on real quantum hardware. Through a structured scoping review, the authors systematically analyze 13 generative systems and five datasets, proposing the first taxonomy based on output modality and training paradigm. They further introduce a three-tiered evaluation framework encompassing syntactic validity, semantic correctness, and hardware executability. The findings reveal that while existing methods generally satisfy syntactic requirements and partially meet semantic criteria, none demonstrate end-to-end validation on actual quantum devices. This critical gap underscores the field’s current limitation in real-device verification and provides a clear direction for future research toward practical, hardware-aware quantum program synthesis.
0 citationsRead paper
Malware Detection based on API Calls: A Reproducibility Study
This study validates and reproduces a malware detection approach based on API call frequency, with a focus on its reproducibility and robustness. Using a Random Forest classifier, the authors construct four model variants—Unigram, Bigram, Trigram, and combined n-grams—and evaluate them on API sequences of length 2,500. The experiments independently confirm for the first time that the original method exhibits high performance stability across different random seeds, with a standard deviation below 0.5% over three trials. Moreover, all model variants achieve F1 scores surpassing those reported in the original study by 0.99% to 2.57%, with the Unigram model reaching an F1 score of 0.8717. These results significantly enhance the scientific rigor and practical viability of the API-based malware detection methodology.
0 citationsRead paper