skopt_gp_minimize_noisy_function_bayesian_optimization.py

1import numpy as np
2from skopt import gp_minimize
3
4def f(x):
5    return (np.sin(5 * x[0]) * (1 - np.tanh(x[0] ** 2)) +
6            np.random.randn() * 0.1)
7
8res = gp_minimize(f,                  # the function to minimize
9                  [(-2.0, 2.0)],      # the bounds on each dimension of x
10                  acq_func="EI",      # the acquisition function
11                  n_calls=15,         # the number of evaluations of f
12                  n_random_starts=5,  # the number of random initialization points
13                  noise=0.1**2,       # the noise level (optional)
14                  random_state=1234)   # the random seed
15
16# Output the best input and the corresponding function value
17print(f"Best score= {res.fun:.4f}")
18print(f"Best parameters= {res.x}")