Efficient initialization for multi-fidelity surrogate-based optimization

The performance of surrogate-based optimization is dependent on the surrogate training set, certainly for realistic optimizations where high cost computing set data imposes small sizes. This especially true multi-fidelity models, different sets exist each fidelity. Adaptive sampling methods have been developed to improve fitting capabilities adding points only necessary or most useful process (i.e., providing highest knowledge gain) and avoiding need an a priori design experiments. Nevertheless, efficiency adaptive highly affected by its initialization. paper presents discusses novel initialization strategy with limited sampling. proposed aims reduce computational evaluating initial set. Furthermore, it allows model adapt more freely data. In this work, approach applied single- stochastic radial basis functions analytical test problem shape NACA hydrofoil. Numerical results show that are improved, thanks effective efficient domain space exploration significant reduction high-fidelity evaluations.