Minion: A C++ and Python Library for Single-Objective Optimization

Optimization problems arise throughout science, engineering, and machine learning, including tasks such as simulation calibration, experimental data fitting, hyperparameter tuning, and design optimization.A common setting is single-objective optimization, where the goal is to minimize or maximize one performance measure.In many practical cases, the objective function is a black box: it can be evaluated at chosen points, but no closed-form expression, gradient, or other derivative information is available.To support optimization problems of this kind, we introduce Minion, a derivative-free optimization library for unconstrained and bound-constrained single-objective black-box problems.It is implemented in C++, with a Python interface available through MinionPy.