Black Box Optimization

Black-box optimization (BBO) tackles the challenge of optimizing functions whose internal workings are unknown, focusing on efficiently finding optimal solutions through iterative evaluations. Current research emphasizes developing more efficient and robust algorithms, including those based on Bayesian optimization, evolutionary strategies, and neural networks (e.g., physics-informed neural networks, transformers), often incorporating techniques like surrogate modeling and multi-fidelity approaches to manage computational costs. BBO's significance lies in its broad applicability across diverse fields, from engineering design and materials science to machine learning hyperparameter tuning and even adversarial attacks on machine learning models, enabling efficient exploration of complex search spaces.