Selected work

Representative Papers

From Model-Based Screening to Data-Driven Surrogates: A Multi-Stage Workflow for Exploring Stochastic Agent-Based Models

Apr 03, 2026

High-dimensional stochastic agent-based models (ABMs) are notoriously difficult to analyze systematically due to the curse of dimensionality and inherent stochasticity. This work proposes a multi-stage automated exploration framework that first employs model-driven experimental design to identify key variables and partition the parameter space, then leverages machine learning surrogate models to efficiently capture residual nonlinear interaction effects. The approach operates without human intervention, automatically detecting unstable regions within the simulator and enabling robust sensitivity analysis and policy testing. Applied to a predator–prey case study, the framework successfully isolates dominant variables and highly sensitive nonlinear regimes, substantially enhancing the efficiency and reliability of ABM exploration.

0 citationsRead paper

Geometric Rough Paths above Mixed Fractional Brownian Motion

Nov 24, 2025

This paper addresses the construction of a geometric rough path theory for mixed fractional Brownian motion (MFBM) and its multicomponent generalization—previously an open problem. Method: We construct, for the first time, the canonical geometric rough path lift of MFBM by approximating it via dyadic smooth paths, employing p-variation estimates, Skorohod integral representations from Malliavin calculus, and Lyons’ universal limit theorem to rigorously handle regularity coupling among components with distinct Hurst exponents. Contribution/Results: Under the minimal condition that the smallest Hurst exponent satisfies (H_{min} > 1/4), we establish existence and uniqueness of the canonical geometric rough path associated with MFBM, thereby ensuring well-posedness of rough differential equations driven by it. Furthermore, we fully characterize the algebraic structure of this lift, unifying the treatment across arbitrary numbers of fractional components and substantially extending the Coutin–Qian theory—originally limited to single-component fractional Brownian motion—to the mixed setting.

0 citationsRead paper

Banks, Bonds, and Collateral: A Microfounded Comparison under Adverse Selection

Sep 03, 2025

This paper investigates firms’ choice between bank loans and corporate bonds under adverse selection. We develop a dynamic contracting model with microfoundations, centered on the key wedge that concentrated creditors (banks) exhibit higher liquidation efficiency than dispersed creditors (bondholders), capturing the fundamental distinction in default resolution: banks enforce contracts more effectively and coordinate more readily, whereas bond markets suffer from coordination frictions and weaker enforcement. Theoretically, we show that coexistence of both debt instruments arises endogenously from creditor structure differences; a sharp financing threshold governs the choice; and testable comparative statics emerge—e.g., improvements in bankruptcy efficiency or bondholder coordination increase bond issuance, narrow the credit spread between safe firms’ bonds and loans, and systematically alter default rates, collateral requirements, and debt maturity composition. All model primitives map directly to observable outcomes—including recovery rates and covenant stringency—enabling welfare decomposition for institutional design.

0 citationsRead paper

Recent publications

Latest Papers

From Model-Based Screening to Data-Driven Surrogates: A Multi-Stage Workflow for Exploring Stochastic Agent-Based Models

Apr 03, 2026

High-dimensional stochastic agent-based models (ABMs) are notoriously difficult to analyze systematically due to the curse of dimensionality and inherent stochasticity. This work proposes a multi-stage automated exploration framework that first employs model-driven experimental design to identify key variables and partition the parameter space, then leverages machine learning surrogate models to efficiently capture residual nonlinear interaction effects. The approach operates without human intervention, automatically detecting unstable regions within the simulator and enabling robust sensitivity analysis and policy testing. Applied to a predator–prey case study, the framework successfully isolates dominant variables and highly sensitive nonlinear regimes, substantially enhancing the efficiency and reliability of ABM exploration.

0 citationsRead paper

Geometric Rough Paths above Mixed Fractional Brownian Motion

Nov 24, 2025

This paper addresses the construction of a geometric rough path theory for mixed fractional Brownian motion (MFBM) and its multicomponent generalization—previously an open problem. Method: We construct, for the first time, the canonical geometric rough path lift of MFBM by approximating it via dyadic smooth paths, employing p-variation estimates, Skorohod integral representations from Malliavin calculus, and Lyons’ universal limit theorem to rigorously handle regularity coupling among components with distinct Hurst exponents. Contribution/Results: Under the minimal condition that the smallest Hurst exponent satisfies (H_{min} > 1/4), we establish existence and uniqueness of the canonical geometric rough path associated with MFBM, thereby ensuring well-posedness of rough differential equations driven by it. Furthermore, we fully characterize the algebraic structure of this lift, unifying the treatment across arbitrary numbers of fractional components and substantially extending the Coutin–Qian theory—originally limited to single-component fractional Brownian motion—to the mixed setting.

0 citationsRead paper

Banks, Bonds, and Collateral: A Microfounded Comparison under Adverse Selection

Sep 03, 2025

This paper investigates firms’ choice between bank loans and corporate bonds under adverse selection. We develop a dynamic contracting model with microfoundations, centered on the key wedge that concentrated creditors (banks) exhibit higher liquidation efficiency than dispersed creditors (bondholders), capturing the fundamental distinction in default resolution: banks enforce contracts more effectively and coordinate more readily, whereas bond markets suffer from coordination frictions and weaker enforcement. Theoretically, we show that coexistence of both debt instruments arises endogenously from creditor structure differences; a sharp financing threshold governs the choice; and testable comparative statics emerge—e.g., improvements in bankruptcy efficiency or bondholder coordination increase bond issuance, narrow the credit spread between safe firms’ bonds and loans, and systematically alter default rates, collateral requirements, and debt maturity composition. All model primitives map directly to observable outcomes—including recovery rates and covenant stringency—enabling welfare decomposition for institutional design.

0 citationsRead paper