Make an option to assume Pareto optimality in WFG

[nabenabe0928]

Motivation

As the current WFG significantly slows down when solution_set is not Pareto optimal, I added a line to make solution_set Pareto optimal.

from optuna._hypervolume import WFG


rng = np.random.RandomState(42)
sols = rng.normal(size=(10000, 3)) + 10.0
reference_point = np.max(sols, axis=0) * 1.1
hv = WFG().compute(sols, reference_point)

As is_pareto_front is not really a cheap operation, I additionally added an option to avoid the function call as well.

Description of the changes

• Add a line to make solution_set Pareto optimal,
• Add an option to avoid the Pareto solution filtering.

[nabenabe0928]

This PR depends on

• Suppress warnings from numpy in hypervolume computation #5432

[github-actions]

This pull request has not seen any recent activity.

[eukaryo]

@nabenabe0928 PR #5432 has been merged!