Numerical Black Box Optimization

Numerical black-box optimization focuses on finding the minimum or maximum of a function whose form is unknown, relying solely on evaluating its output for given inputs. Current research emphasizes improving existing algorithms like CMA-ES and Differential Evolution, particularly addressing challenges posed by noise, mixed-integer variables, and high dimensionality, often through adaptive parameter control and algorithm switching strategies. These advancements are crucial for tackling complex real-world problems across diverse fields, from engineering design to machine learning, where analytical function descriptions are unavailable or computationally intractable.