Algorithm 1: Bayesian Hyperparameter Optimization

Input:

Initial observation set $D_n=\{x_n,y_n\}$

Bounds for the search space $\chi$

Output: ${\{x_n,y_n\}}_{n=1}^t$

for$n=1,2,\dots ,$ t do

Fit the current data sample $D_n$ to get the GPR model $G{(w)}_n$

Solve the extreme points of the objective function $loss(w)$:$w_{n+1}=\underset{w\in W}{\arg \min }loss(w,G{(w)}_n)$

Obtain new samples $\left(w_{n+1},loss{\left(w\right)}_{n+1}\right)$

Update data sample $D_{n+1}=\left\{D_n\left(w_{n+1},loss{\left(w\right)}_{n+1}\right)\right\}$.

Update data self-screening layer parameters

end for