Bayesian optimizations. The optimization processes consider all kernels and basis functions listed in Eqs. (1)–(14). The “nonisotropic exponential” kernel (Eq. (6)) and “linear” basis function (Eq. (13)) are selected by both the LCB and EIPSP algorithms. “Sigma” here is used as the initial value in building models 1 and 2. CV Fi (i = 1, 2, …, 20) is listed in Table 1. (a) LCB. (b) EIPSP.