Mathematics & Theoretical Foundations
Yaprak Güldoğan Dericioğlu
Abstract
The Extended Block Arnoldi–Backward Differentiation Formula (EBA–BDF) is a projection-based integrator for solving large-scale Matrix Differential Riccati Equations (DREs). Like many Krylov subspace methods, its performance depends on the choice of the subspace dimension. In practice, this ...
Read More
The Extended Block Arnoldi–Backward Differentiation Formula (EBA–BDF) is a projection-based integrator for solving large-scale Matrix Differential Riccati Equations (DREs). Like many Krylov subspace methods, its performance depends on the choice of the subspace dimension. In practice, this parameter is often determined through empirical tuning. In this work, we introduce a lightweight, data-driven pre-solver to estimate this dimension \emph{a priori}. The approach uses a Random Forest model trained on spectral norms and discretization parameters, and predicts the required subspace size without modifying the numerical core or stability properties of the original method. Numerical experiments show that the proposed approach can automate parameter selection and reduce the need for manual tuning. The effect is more noticeable in diffusion-dominated regimes, where spectral properties lead to more regular Krylov convergence. By simplifying the initialization stage, the approach supports the practical use of EBA–BDF solvers in large-scale problems.
Mathematics & Theoretical Foundations
Saad Qasim Abbas; Wasan Saad Ahmed
Abstract
This paper presents a systematic comparative study of two widely used numerical solvers --- HOFiD_bvp (high-order finite difference scheme) and bvp4c (collocation-based) --- for solving singular second-order ordinary differential equations (ODEs) with first-kind (regular) boundary singularities. Four ...
Read More
This paper presents a systematic comparative study of two widely used numerical solvers --- HOFiD_bvp (high-order finite difference scheme) and bvp4c (collocation-based) --- for solving singular second-order ordinary differential equations (ODEs) with first-kind (regular) boundary singularities. Four representative benchmark problems drawn from fluid dynamics, materials science, and radially symmetric diffusion models are used to evaluate solver performance across key metrics: maximum residual, maximum error, mesh point count, and ODE/BC function call counts. Results show that HOFiD_bvp consistently achieves lower residuals and errors with fewer function evaluations, making it computationally more efficient. Conversely, bvp4c demonstrates superior robustness for nonlinear singular problems and offers better adaptive mesh refinement capabilities. These findings provide practical guidance for selecting the appropriate numerical technique in applied science and engineering contexts, with implications for optimization of computational simulation workflows.
