A Derivative-free Two Level Random Search Method for Unconstrained Optimization

Focusing on a novel derivative-free optimization method, this book introduces an algorithm that utilizes randomly generated trial points within defined domains. At each iteration, the best points are selected based on various rules, setting it apart from traditional methods. The approach demonstrates effectiveness in tackling a wide range of unconstrained optimization problems, particularly those with high variable counts, as evidenced by extensive numerical experiments involving 140 problems with up to 500 variables, showcasing its efficiency and robustness.