What is safeopt?

SafeOpt is a Python library for optimizing expensive black-box functions (i.e., functions for which the mathematical form is unknown) with safety constraints. It uses a Bayesian optimization approach, specifically an efficient global optimization (EGO) algorithm, to iteratively evaluate the expensive function and update a probabilistic model of the function. The model is used to guide the search towards regions of the input space that are likely to contain a global minimum, while the safety constraints are used to ensure that any candidate solutions that are found are safe to use. SafeOpt is particularly useful in applications where the cost of evaluating the function is high, such as in engineering design or medical applications.