Box Optimization

Box optimization, encompassing both black-box and gray-box approaches, aims to efficiently find optimal solutions for complex problems where evaluating the objective function is computationally expensive. Current research focuses on leveraging partial evaluations (gray-box) within algorithms like GOMEA and Bayesian Optimization, incorporating problem-specific knowledge to improve efficiency through techniques such as linkage learning and conditional modeling, and parallelization strategies like GPU acceleration. These advancements significantly enhance the speed and scalability of optimization, impacting diverse fields from sensor placement in assisted living to solving combinatorial problems, by reducing the number of expensive function evaluations needed to find high-quality solutions.