Key research themes
1. How can the Central Difference Method be adapted for solving nonlinear least squares and nonlinear systems without explicit Jacobian computations?
This research area focuses on iterative numerical schemes that utilize central difference or divided difference approximations to replace explicit Jacobians or derivatives in solving nonlinear least squares problems and nonlinear equations. It is particularly important when derivatives are difficult, expensive, or impossible to compute exactly, and when the residual or objective function contains both differentiable and nondifferentiable parts. Such methods strive to maintain high convergence order and numerical efficiency without sacrificing robustness.
2. What are the strategies and theoretical guarantees for constructing derivative-free or low-derivative iterative methods with high-order convergence using central difference approximations?
This theme explores approaches to build high-order root-finding or nonlinear equation solvers that avoid explicit derivative computations through central difference or finite difference approximations. It covers theoretical convergence analyses of these derivative-free schemes, including multipoint and multi-step methods that achieve quadratic or higher order convergence. The motivation arises from scenarios where derivatives are unavailable, expensive, or unreliable, and these methods provide efficient alternatives while maintaining competitive convergence speed.
3. How can the classical Central Difference Method be extended or modified for applications to partial differential equations and matrix problems, including interval arithmetic adaptations?
This research direction concerns advancing the classical central difference finite difference scheme to novel problem domains such as Poisson equations and matrix adjustment tasks, including the use of interval arithmetic to rigorously bound errors introduced by floating point, discretization, and method approximations. It emphasizes method modifications to guarantee inclusion of exact solutions in interval enclosures and improved stability of solutions for PDEs and matrix estimation problems, which is critical for robust scientific computing and engineering applications.