Selected work

Representative Papers

Strong imposition of Dirichlet boundary velocities in structure-preserving discretizations of elastodynamics

Jul 28, 2026

This work addresses the challenge of enforcing Dirichlet boundary velocities in structure-preserving discretizations, which traditionally either introduce Lagrange multipliers—yielding differential-algebraic equations—or weakly impose boundary conditions at the expense of accuracy. The authors propose a continuous-level additive kinematic decomposition that splits the displacement and velocity fields into a dynamic component satisfying homogeneous boundary conditions and a prescribed lifting function. Building on the principle of virtual power, they formulate a lifted port-Hamiltonian system. Upon finite element discretization, the resulting system is an ordinary differential equation that strongly enforces Dirichlet velocity boundary conditions—a first within the port-Hamiltonian framework—while preserving energy conservation and the ODE structure. The approach also unifies classical finite element matrix partitioning techniques. Numerical experiments confirm its energy balance, computational efficiency, and equivalence to standard formulations.

0 citationsRead paper

Dirichlet-Based Monte Carlo Dropout for Uncertainty Estimation in Neural Networks

May 22, 2026

Traditional neural networks struggle to provide reliable uncertainty estimates, while Bayesian neural networks, despite their theoretical advantages, are computationally expensive and difficult to scale. This work proposes a novel approach that integrates Dirichlet distributions with Monte Carlo Dropout to structurally model predictive class probabilities, thereby yielding better-calibrated and more informative uncertainty representations while retaining the computational efficiency of Monte Carlo Dropout during inference. The method demonstrates significant improvements over existing techniques across multiple benchmark tasks, offering a practical pathway toward deploying efficient and reliable uncertainty-aware deep models in real-world applications.

0 citationsRead paper

Hamming distance between finite transducers

Apr 28, 2026

This study addresses the bounded discrepancy problem for two nondeterministic finite transducers, asking whether the Hamming distance between their output strings is at most a given threshold \(k\) for all inputs. Employing techniques from formal language theory, automata theory, and computational complexity, and via logspace many-one reductions, the work establishes the precise complexity classification of this problem for the first time: it is NL-complete when \(k\) is fixed, co-NP-complete when \(k\) is given in binary, and DP-complete when the Hamming distance is required to be exactly \(k\). Furthermore, the paper proves that the maximum Hamming distance between two such transducers admits a tight quadratic upper bound, which is asymptotically optimal with respect to the size of the transducers.

0 citationsRead paper

Decomposition of Automata recognizing Ideals

Apr 28, 2026

This study addresses the decomposability of automata recognizing ideal languages in level 1/2 of the Straubing–Thérien hierarchy, specifically investigating whether such automata can be expressed as intersections or unions of smaller automata. For the first time, we establish that the corresponding decomposition decision problem lies within the complexity class NL. Furthermore, we propose a polynomial-time algorithm that effectively computes intersection decompositions while preserving the ideality of the language. By integrating techniques from formal language theory with structural analysis of automata, our approach enables efficient and property-preserving decomposition of ideal-language automata, significantly enhancing their tractability and modularity.

0 citationsRead paper

Risk Is Not the Target: A Monotonic Framework for Evaluating Wildfire Operational Risk Signals

Apr 23, 2026

Traditional wildfire risk assessments rely heavily on classification metrics such as F1 score and Intersection over Union (IoU), often overlooking the monotonic relationship between risk scores and operational burdens—such as actual fire counts and resource allocation. This work proposes a novel monotonicity-based evaluation framework that shifts the focus of risk model assessment from mere predictive accuracy to explanatory power regarding operational dynamics. The framework is applied to compare the expert-derived DFE index, a GRU-based temporal model, and the multi-agent system FARS in the Alpes-Maritimes region of France. Results reveal that while DFE exhibits weaker classification performance, it achieves the strongest global monotonicity; GRU demonstrates strong local monotonicity but uneven risk distribution; and FARS exposes structural deficiencies in upstream signal processing. These findings validate the efficacy of an evaluation paradigm centered on operational dynamics.

0 citationsRead paper

Recent publications

Latest Papers

Strong imposition of Dirichlet boundary velocities in structure-preserving discretizations of elastodynamics

Jul 28, 2026

This work addresses the challenge of enforcing Dirichlet boundary velocities in structure-preserving discretizations, which traditionally either introduce Lagrange multipliers—yielding differential-algebraic equations—or weakly impose boundary conditions at the expense of accuracy. The authors propose a continuous-level additive kinematic decomposition that splits the displacement and velocity fields into a dynamic component satisfying homogeneous boundary conditions and a prescribed lifting function. Building on the principle of virtual power, they formulate a lifted port-Hamiltonian system. Upon finite element discretization, the resulting system is an ordinary differential equation that strongly enforces Dirichlet velocity boundary conditions—a first within the port-Hamiltonian framework—while preserving energy conservation and the ODE structure. The approach also unifies classical finite element matrix partitioning techniques. Numerical experiments confirm its energy balance, computational efficiency, and equivalence to standard formulations.

0 citationsRead paper

Dirichlet-Based Monte Carlo Dropout for Uncertainty Estimation in Neural Networks

May 22, 2026

Traditional neural networks struggle to provide reliable uncertainty estimates, while Bayesian neural networks, despite their theoretical advantages, are computationally expensive and difficult to scale. This work proposes a novel approach that integrates Dirichlet distributions with Monte Carlo Dropout to structurally model predictive class probabilities, thereby yielding better-calibrated and more informative uncertainty representations while retaining the computational efficiency of Monte Carlo Dropout during inference. The method demonstrates significant improvements over existing techniques across multiple benchmark tasks, offering a practical pathway toward deploying efficient and reliable uncertainty-aware deep models in real-world applications.

0 citationsRead paper

Hamming distance between finite transducers

Apr 28, 2026

This study addresses the bounded discrepancy problem for two nondeterministic finite transducers, asking whether the Hamming distance between their output strings is at most a given threshold \(k\) for all inputs. Employing techniques from formal language theory, automata theory, and computational complexity, and via logspace many-one reductions, the work establishes the precise complexity classification of this problem for the first time: it is NL-complete when \(k\) is fixed, co-NP-complete when \(k\) is given in binary, and DP-complete when the Hamming distance is required to be exactly \(k\). Furthermore, the paper proves that the maximum Hamming distance between two such transducers admits a tight quadratic upper bound, which is asymptotically optimal with respect to the size of the transducers.

0 citationsRead paper

Decomposition of Automata recognizing Ideals

Apr 28, 2026

This study addresses the decomposability of automata recognizing ideal languages in level 1/2 of the Straubing–Thérien hierarchy, specifically investigating whether such automata can be expressed as intersections or unions of smaller automata. For the first time, we establish that the corresponding decomposition decision problem lies within the complexity class NL. Furthermore, we propose a polynomial-time algorithm that effectively computes intersection decompositions while preserving the ideality of the language. By integrating techniques from formal language theory with structural analysis of automata, our approach enables efficient and property-preserving decomposition of ideal-language automata, significantly enhancing their tractability and modularity.

0 citationsRead paper

Risk Is Not the Target: A Monotonic Framework for Evaluating Wildfire Operational Risk Signals

Apr 23, 2026

Traditional wildfire risk assessments rely heavily on classification metrics such as F1 score and Intersection over Union (IoU), often overlooking the monotonic relationship between risk scores and operational burdens—such as actual fire counts and resource allocation. This work proposes a novel monotonicity-based evaluation framework that shifts the focus of risk model assessment from mere predictive accuracy to explanatory power regarding operational dynamics. The framework is applied to compare the expert-derived DFE index, a GRU-based temporal model, and the multi-agent system FARS in the Alpes-Maritimes region of France. Results reveal that while DFE exhibits weaker classification performance, it achieves the strongest global monotonicity; GRU demonstrates strong local monotonicity but uneven risk distribution; and FARS exposes structural deficiencies in upstream signal processing. These findings validate the efficacy of an evaluation paradigm centered on operational dynamics.

0 citationsRead paper