Deep Hedging Under Realistic Market Frictions: A Regime-Conditional Empirical Study of Dynamic Option Hedging on Bitcoin Options
研究使用实际比特币期权数据测试了深度对冲与经典对冲方法在市场摩擦下的表现，发现Whalley-Wilmott方法在降低成本方面优于其他方法，而深度学习模型未表现出优势。
0 citationsRead paper
Selected work
Representative Papers
Queue occupancy and server size distribution of a queue length dependent vacation queue with an optional service
This study addresses the joint optimization of energy efficiency and quality of service under bursty traffic in emerging scenarios such as 5G and IoT by analyzing a discrete-time queue with batch arrivals, infinite buffer capacity, two-phase service (mandatory followed by optional), and single/multiple server vacations. By constructing a bivariate probability generating function, the work fully characterizes—for the first time—the joint steady-state distribution of queue length and server status at an arbitrary time slot, including vacation termination instants, and integrates optional service with a queue-length-dependent vacation policy. Leveraging discrete-phase-type distribution modeling and sensitivity analysis, the authors derive closed-form expressions for key system performance metrics. Numerical experiments validate the analytical model and elucidate the impact of critical parameters on system behavior.
0 citationsRead paper
PSSI-MaxST: An Efficient Pixel-Segment Similarity Index Using Intensity and Smoothness Features for Maximum Spanning Tree Based Segmentation
This work proposes an efficient interactive image segmentation framework based on a Pixel-Superpixel Similarity Index (PSSI) to address the limitations of existing methods, which often suffer from high computational cost, sensitivity to user input, and performance degradation when foreground and background exhibit similar colors. The approach leverages MeanShift to generate initial superpixels, constructs a pixel-superpixel graph, and employs a Maximum Spanning Tree (MaxST) for segmentation, jointly modeling color, texture, shape, and local strong connectivity. A key innovation lies in computing PSSI via the harmonic mean of multi-channel similarities, effectively mitigating inconsistencies from individual channels and enhancing robustness. Notably, PSSI achieves a computational complexity of only O(B), significantly improving efficiency. Experimental results on the GrabCut and Images250 datasets demonstrate superior performance over state-of-the-art methods—including AMOE, OneCut, and SSNCut—in terms of Jaccard index, F1 score, execution time, and average error.
0 citationsRead paper
Higher-order spacings in the superposed spectra of random matrices with comparison to spacing ratios and application to complex systems
This study investigates higher-order spectral spacing statistics (k ≥ 2) of superimposed random matrices to characterize spectral fluctuations and symmetry structures in complex quantum systems. Method: We perform extensive numerical simulations on circular orthogonal/ensemble (COE/GOE) ensembles and paradigmatic quantum chaotic models—the quantum kicked top and intermediate map—systematically analyzing asymptotic behavior and finite-size corrections of higher-order spacing distributions and spacing ratios. Contribution/Results: We introduce a modified Dyson index β′, enabling direct symmetry-class identification without spectral unfolding. We find that for m ≥ 2, higher-order spacing and ratio statistics become asymptotically identical, yet significant deviations persist for small k. Moreover, for k = 1, spacing ratios converge to the Poisson limit faster than spacings as m increases. We further quantify how the randomness of generating matrices influences higher-order spectral statistics. These results provide novel, empirically validated criteria for symmetry identification and universality in quantum chaos.
0 citationsRead paper
Lightweight Clinical Decision Support System using QLoRA-Fine-Tuned LLMs and Retrieval-Augmented Generation
Large language models (LLMs) deployed in clinical decision support systems (CDSS) face critical bottlenecks in accuracy, inference efficiency, and interpretability—especially under resource-constrained, privacy-sensitive, and regulatory-compliant clinical environments. Method: We propose a lightweight, domain-adapted architecture integrating QLoRA fine-tuning with medical-domain-specific retrieval-augmented generation (RAG). Built upon Llama 3.2-3B-Instruct, it incorporates hospital-specific clinical data, ontology-aligned medical knowledge graphs, and FAISS/Chroma-based vector retrieval, coupled with 4-bit NF4 quantization for high-fidelity model compression. Contribution/Results: Our novel QLoRA-RAG co-design achieves parameter efficiency (75% GPU memory reduction), low-latency inference (<800 ms), offline edge deployability, and clinically verifiable outputs. Evaluated on multiple medical benchmarks, it attains F1 ≥ 0.82—meeting clinical usability thresholds—and significantly improves accuracy, privacy preservation, and scalability across disease prediction, treatment recommendation, and clinical note summarization tasks.
0 citationsRead paper
Recent publications
Latest Papers
Deep Hedging Under Realistic Market Frictions: A Regime-Conditional Empirical Study of Dynamic Option Hedging on Bitcoin Options
研究使用实际比特币期权数据测试了深度对冲与经典对冲方法在市场摩擦下的表现，发现Whalley-Wilmott方法在降低成本方面优于其他方法，而深度学习模型未表现出优势。
0 citationsRead paper
Queue occupancy and server size distribution of a queue length dependent vacation queue with an optional service
This study addresses the joint optimization of energy efficiency and quality of service under bursty traffic in emerging scenarios such as 5G and IoT by analyzing a discrete-time queue with batch arrivals, infinite buffer capacity, two-phase service (mandatory followed by optional), and single/multiple server vacations. By constructing a bivariate probability generating function, the work fully characterizes—for the first time—the joint steady-state distribution of queue length and server status at an arbitrary time slot, including vacation termination instants, and integrates optional service with a queue-length-dependent vacation policy. Leveraging discrete-phase-type distribution modeling and sensitivity analysis, the authors derive closed-form expressions for key system performance metrics. Numerical experiments validate the analytical model and elucidate the impact of critical parameters on system behavior.
0 citationsRead paper
PSSI-MaxST: An Efficient Pixel-Segment Similarity Index Using Intensity and Smoothness Features for Maximum Spanning Tree Based Segmentation
This work proposes an efficient interactive image segmentation framework based on a Pixel-Superpixel Similarity Index (PSSI) to address the limitations of existing methods, which often suffer from high computational cost, sensitivity to user input, and performance degradation when foreground and background exhibit similar colors. The approach leverages MeanShift to generate initial superpixels, constructs a pixel-superpixel graph, and employs a Maximum Spanning Tree (MaxST) for segmentation, jointly modeling color, texture, shape, and local strong connectivity. A key innovation lies in computing PSSI via the harmonic mean of multi-channel similarities, effectively mitigating inconsistencies from individual channels and enhancing robustness. Notably, PSSI achieves a computational complexity of only O(B), significantly improving efficiency. Experimental results on the GrabCut and Images250 datasets demonstrate superior performance over state-of-the-art methods—including AMOE, OneCut, and SSNCut—in terms of Jaccard index, F1 score, execution time, and average error.
0 citationsRead paper
Higher-order spacings in the superposed spectra of random matrices with comparison to spacing ratios and application to complex systems
This study investigates higher-order spectral spacing statistics (k ≥ 2) of superimposed random matrices to characterize spectral fluctuations and symmetry structures in complex quantum systems. Method: We perform extensive numerical simulations on circular orthogonal/ensemble (COE/GOE) ensembles and paradigmatic quantum chaotic models—the quantum kicked top and intermediate map—systematically analyzing asymptotic behavior and finite-size corrections of higher-order spacing distributions and spacing ratios. Contribution/Results: We introduce a modified Dyson index β′, enabling direct symmetry-class identification without spectral unfolding. We find that for m ≥ 2, higher-order spacing and ratio statistics become asymptotically identical, yet significant deviations persist for small k. Moreover, for k = 1, spacing ratios converge to the Poisson limit faster than spacings as m increases. We further quantify how the randomness of generating matrices influences higher-order spectral statistics. These results provide novel, empirically validated criteria for symmetry identification and universality in quantum chaos.
0 citationsRead paper
Lightweight Clinical Decision Support System using QLoRA-Fine-Tuned LLMs and Retrieval-Augmented Generation
Large language models (LLMs) deployed in clinical decision support systems (CDSS) face critical bottlenecks in accuracy, inference efficiency, and interpretability—especially under resource-constrained, privacy-sensitive, and regulatory-compliant clinical environments. Method: We propose a lightweight, domain-adapted architecture integrating QLoRA fine-tuning with medical-domain-specific retrieval-augmented generation (RAG). Built upon Llama 3.2-3B-Instruct, it incorporates hospital-specific clinical data, ontology-aligned medical knowledge graphs, and FAISS/Chroma-based vector retrieval, coupled with 4-bit NF4 quantization for high-fidelity model compression. Contribution/Results: Our novel QLoRA-RAG co-design achieves parameter efficiency (75% GPU memory reduction), low-latency inference (<800 ms), offline edge deployability, and clinically verifiable outputs. Evaluated on multiple medical benchmarks, it attains F1 ≥ 0.82—meeting clinical usability thresholds—and significantly improves accuracy, privacy preservation, and scalability across disease prediction, treatment recommendation, and clinical note summarization tasks.
0 citationsRead paper