Algorithm 1: HORD Algorithm.

1: Generate a little insight $A_{t0}= {\left\{h^{\left(d\right)}, G\left(h^{\left(d\right)}\right)\right\}}_{d=1}^{t_0}$ measuring with Latin hypercubes;

2: while $t <N_{max}$ do

3: Fit or revise the$S_t\left(h\right)$RBF interpolation model from (4) using$A_t$.

4: Fix $h_{test}=argmax\left\{G\left(h\right)\right\} in A_t$ ;

5: Create c candidates using $h_{cand}^{\left(1:c\right)}$ based on $h_{test}$ and ${\delta }_d$ samples were taken from a normal distribution with a certain probability ${\phi }_t$ in (5).;

6: Calculate $W_t^{cv}{h}_{cand}^{\left(1:c\right)}$ by (7), $W_t^{dm}{h}_{cand}^{\left(1:c\right)} by \left(8\right)$ , as well as the final weighted score $W_t{h}_{cand}^{\left(1:c\right)}$ by (6).

7: Set $h^{*}=argmin\left\{W_t{h}_{cand}^{\left(1:c\right)}\right\}$ .

8: Estimate $G\left(h^{*}\right)$ .

9: Update $A_{t+1}= \left\{A_t \cup \left(h^{*}, G\left(h^{*}\right)\right)\right\}$

10: end while

11: Find$h_{test}=argmax\left\{G\left(h\right)\right\} in A_{N_{max}}$.

12: Return$h_{test}$.