Abstract
Several numerical differential equation solvers have been employedeffectively over the years as an alternative to analytical solvers to quicklyand conveniently solve differential equations. One category of these isboundary value solvers, which are used to solve real-world problems formulatedas differential equations with boundary conditions. These solvers requirecertain numerical settings to solve the differential equations that affecttheir solvability and performance. A systematic fine-tuning of these settingsis required to obtain the desired solution and performance. Currently, thesesettings are either selected by trial and error or require domain expertise. Inthis paper, we propose a machine learning-based optimization workflow forfine-tuning the numerical settings to reduce the time and domain expertiserequired in the process. In the evaluation section, we discuss the scalability,stability, and reliability of the proposed workflow. We demonstrate ourworkflow on a numerical boundary value problem solver.