Selected work

Representative Papers

A perspective note on likelihood approximation and inference for complex simulation models using a chain of aggregated normalizing flows

Oct 05, 2026

This study addresses the scalability and statistical validity bottlenecks in likelihood approximation and inference for complex simulation models by proposing a novel framework based on aggregated normalizing flow chains. Methodologically, it integrates information-theoretic formalization with sequential decision-making paradigms to construct flexible probability distributions through the sequential optimization of bijective transformation parameters. Furthermore, an empirical likelihood estimator under moment constraints is employed to iteratively update and aggregate the global flow parameters. This research establishes a surrogate model that simultaneously ensures computational feasibility and statistical power, enabling efficient parameter exploration, hypothesis testing, and uncertainty quantification. Ultimately, the proposed approach provides a reliable Bayesian inference solution for complex systems.

0 citationsRead paper

Simulation-based parameter estimation via a combination of embedded normalizing flows and implied empirical probabilities under moment restrictions

Jul 27, 2026

This work addresses the challenge of parameter estimation in physical system simulation models arising from complex residual distributions. The authors propose an end-to-end estimation framework that employs embedded normalizing flows to map intricate residuals onto a simple base distribution. Indirect constraints are imposed on this base distribution through empirical likelihood under moment conditions. Model and flow parameters are jointly optimized via implicit differentiation combined with first-order gradient methods. By innovatively integrating normalizing flows with constrained empirical likelihood, the approach establishes an information-theoretically interpretable and computationally tractable framework. The resulting inverse transformation serves as an invertible surrogate model, enhancing both the accuracy and efficiency of parameter estimation while enabling quantification of model bias and sensitivity analysis.

0 citationsRead paper

Simulation-based Bayesian inference with ameliorative learned summary statistics -- Part I

Jan 30, 2026

This work proposes a simulation-based Bayesian inference framework to address the challenges posed by intractable or computationally prohibitive likelihood functions. By learning efficient summary statistics to construct an empirical likelihood and incorporating the Cressie–Read divergence criterion under moment constraints, the method transforms simulated data to enable conditioning on observed data. The approach achieves both statistical efficiency and valid conditional inference, naturally accommodating weakly dependent data and distributed computing environments. The resulting unified framework is readily extensible to complex simulator-based models and large-scale data settings, facilitating accurate and computationally efficient Bayesian inference even when the likelihood is unavailable.

0 citationsRead paper

A brief note on learning problem with global perspectives

Jan 09, 2026

This work addresses the challenges in dynamic optimization within leader-follower learning settings, where followers lack a global perspective and the leader struggles to balance private data utility with follower generalization performance. To this end, we propose a novel leader-follower learning framework that introduces, for the first time, a global perspective mechanism enabling followers to perform collaborative learning based on aggregated information shared by the leader. The leader optimizes its own objective and guides follower behavior through a unified formulation that solves a higher-order empirical likelihood estimation problem subject to conditional moment constraints. By integrating empirical likelihood, conditional moment restrictions, and leader-follower game dynamics, the proposed method not only provides a rigorous mathematical characterization of the learning process but also establishes a solid foundation for theoretical analysis.

0 citationsRead paper

WoundNet-Ensemble: A Novel IoMT System Integrating Self-Supervised Deep Learning and Multi-Model Fusion for Automated, High-Accuracy Wound Classification and Healing Progression Monitoring

Dec 20, 2025

Chronic wounds (e.g., diabetic foot ulcers) suffer from subjective clinical assessment and inconsistent classification, leading to delayed interventions. To address this, we propose an Internet-of-Medical-Things (IoMT) system for intelligent chronic wound management. Our method introduces the first weighted multimodal architecture integrating ResNet-50, self-supervised DINOv2 Vision Transformer, and Swin Transformer, coupled with a longitudinal healing tracking mechanism enabling automatic six-class wound classification and dynamic healing quantification—including healing rate estimation and early anomaly detection. The system leverages self-supervised pretraining, multi-scale feature modeling, and edge–cloud collaborative deployment. Evaluated on a dataset of 5,175 multi-etiology wound images, it achieves 99.90% classification accuracy—surpassing state-of-the-art methods by 3.7%. It supports real-time healing scoring, severity quantification, and clinically actionable alerts.

0 citationsRead paper

Recent publications

Latest Papers

A perspective note on likelihood approximation and inference for complex simulation models using a chain of aggregated normalizing flows

Oct 05, 2026

This study addresses the scalability and statistical validity bottlenecks in likelihood approximation and inference for complex simulation models by proposing a novel framework based on aggregated normalizing flow chains. Methodologically, it integrates information-theoretic formalization with sequential decision-making paradigms to construct flexible probability distributions through the sequential optimization of bijective transformation parameters. Furthermore, an empirical likelihood estimator under moment constraints is employed to iteratively update and aggregate the global flow parameters. This research establishes a surrogate model that simultaneously ensures computational feasibility and statistical power, enabling efficient parameter exploration, hypothesis testing, and uncertainty quantification. Ultimately, the proposed approach provides a reliable Bayesian inference solution for complex systems.

0 citationsRead paper

Simulation-based parameter estimation via a combination of embedded normalizing flows and implied empirical probabilities under moment restrictions

Jul 27, 2026

This work addresses the challenge of parameter estimation in physical system simulation models arising from complex residual distributions. The authors propose an end-to-end estimation framework that employs embedded normalizing flows to map intricate residuals onto a simple base distribution. Indirect constraints are imposed on this base distribution through empirical likelihood under moment conditions. Model and flow parameters are jointly optimized via implicit differentiation combined with first-order gradient methods. By innovatively integrating normalizing flows with constrained empirical likelihood, the approach establishes an information-theoretically interpretable and computationally tractable framework. The resulting inverse transformation serves as an invertible surrogate model, enhancing both the accuracy and efficiency of parameter estimation while enabling quantification of model bias and sensitivity analysis.

0 citationsRead paper

Simulation-based Bayesian inference with ameliorative learned summary statistics -- Part I

Jan 30, 2026

This work proposes a simulation-based Bayesian inference framework to address the challenges posed by intractable or computationally prohibitive likelihood functions. By learning efficient summary statistics to construct an empirical likelihood and incorporating the Cressie–Read divergence criterion under moment constraints, the method transforms simulated data to enable conditioning on observed data. The approach achieves both statistical efficiency and valid conditional inference, naturally accommodating weakly dependent data and distributed computing environments. The resulting unified framework is readily extensible to complex simulator-based models and large-scale data settings, facilitating accurate and computationally efficient Bayesian inference even when the likelihood is unavailable.

0 citationsRead paper

A brief note on learning problem with global perspectives

Jan 09, 2026

This work addresses the challenges in dynamic optimization within leader-follower learning settings, where followers lack a global perspective and the leader struggles to balance private data utility with follower generalization performance. To this end, we propose a novel leader-follower learning framework that introduces, for the first time, a global perspective mechanism enabling followers to perform collaborative learning based on aggregated information shared by the leader. The leader optimizes its own objective and guides follower behavior through a unified formulation that solves a higher-order empirical likelihood estimation problem subject to conditional moment constraints. By integrating empirical likelihood, conditional moment restrictions, and leader-follower game dynamics, the proposed method not only provides a rigorous mathematical characterization of the learning process but also establishes a solid foundation for theoretical analysis.

0 citationsRead paper

WoundNet-Ensemble: A Novel IoMT System Integrating Self-Supervised Deep Learning and Multi-Model Fusion for Automated, High-Accuracy Wound Classification and Healing Progression Monitoring

Dec 20, 2025

Chronic wounds (e.g., diabetic foot ulcers) suffer from subjective clinical assessment and inconsistent classification, leading to delayed interventions. To address this, we propose an Internet-of-Medical-Things (IoMT) system for intelligent chronic wound management. Our method introduces the first weighted multimodal architecture integrating ResNet-50, self-supervised DINOv2 Vision Transformer, and Swin Transformer, coupled with a longitudinal healing tracking mechanism enabling automatic six-class wound classification and dynamic healing quantification—including healing rate estimation and early anomaly detection. The system leverages self-supervised pretraining, multi-scale feature modeling, and edge–cloud collaborative deployment. Evaluated on a dataset of 5,175 multi-etiology wound images, it achieves 99.90% classification accuracy—surpassing state-of-the-art methods by 3.7%. It supports real-time healing scoring, severity quantification, and clinically actionable alerts.

0 citationsRead paper