Surrogate Model

A surrogate model serves as a computationally efficient approximation of a more complex and expensive simulation model. In the context of machine learning and optimization, surrogate models are often employed to replace expensive simulations, such as those found in engineering or scientific applications, where evaluations of the true model are costly in terms of time and computational resources. These models are typically constructed using techniques such as Gaussian processes, polynomial regression, or neural networks, which learn the underlying function mapping inputs to outputs based on a limited set of sampled data points. Mathematically, surrogate models can be viewed as a function f: X → Y, where X is the input space and Y is the output space, with the goal of minimizing the expected prediction error over the input distribution. The relationship to parent concepts lies in the broader field of optimization and Bayesian optimization, where surrogate models are used to guide the search for optimal solutions by balancing exploration and exploitation of the input space, thereby reducing the number of costly evaluations needed to find an optimal solution.