"""This module demonstrates the process for a blackbox bayesian optimization using the
Yotse framework."""
from bayes_opt import UtilityFunction
from examples.wobbly_function.example_wobbly_main import remove_files_after_run
from examples.wobbly_function.example_wobbly_main import wobbly_pre
from yotse.execution import Executor
from yotse.pre import OptimizationInfo
[docs]
def main() -> None:
"""Executes a blackbox optimization experiment for a 'wobbly function' of unknown
form (at least we pretend to not know for the purpose of this example).
This function first adds the bayesian optimization information, sets it to active
and finally runs the optimization.
Note: in the current setting we are by default giving the initial datapoints of the experiment to the optimization.
So the first step will execute multiple instances of the optimization while subsequent steps will execute a single
point suggested by the optimization algorithm.
"""
print("\033[93m --- Executing Wobbly-Bayesian Example. --- \033[0m")
bayesopt_experiment = wobbly_pre()
bayes_opt = OptimizationInfo(
name="bayesopt",
blackbox_optimization=True,
opt_parameters={
"utility_function": UtilityFunction(kind="ucb", kappa=2.5, xi=0.0),
"n_iter": 10,
},
is_active=True,
)
# deactivate other optimization
bayesopt_experiment.opt_info_list[0].is_active = False
# append new active whitebox optimization
bayesopt_experiment.opt_info_list.append(bayes_opt)
wobbly_bayesopt = Executor(experiment=bayesopt_experiment)
for i in range(wobbly_bayesopt.optimizer.optimization_algorithm.max_iterations):
wobbly_bayesopt.run(step_number=i)
solution = wobbly_bayesopt.optimizer.suggest_best_solution()
print("Solution: ", solution)
# matplotlib.use('Qt5Agg')
# wobbly_example.optimization_alg.ga_instance.plot_new_solution_rate()
# wobbly_example.optimization_alg.ga_instance.plot_fitness()
remove_files_after_run()
if __name__ == "__main__":
main()