Calibrated uncertainties improve the performance of Bayesian optimization by better balancing exploration and exploitation.
