def _make_pruner(self):
if isinstance(self.pruner_method, str):
if self.pruner_method == 'halving':
pruner = SuccessiveHalvingPruner(
min_resource=self.n_timesteps // 6,
reduction_factor=4,
min_early_stopping_rate=0)
elif self.pruner_method == 'median':
pruner = MedianPruner(n_startup_trials=5,
n_warmup_steps=self.n_timesteps // 6)
elif self.pruner_method == 'none':
# Do not prune
pruner = NopPruner()
else:
raise ValueError('Unknown pruner: {}'.format(
self.pruner_method))
elif isinstance(self.pruner_method, dict):
method_copy = deepcopy(self.pruner_method)
method = method_copy.pop('method')
if method == 'halving':
pruner = SuccessiveHalvingPruner(**method_copy)
elif method == 'median':
pruner = MedianPruner(**method_copy)
elif method == 'none':
# Do not prune
pruner = NopPruner()
else:
raise ValueError('Unknown pruner: {}'.format(
self.pruner_method))
else:
raise ValueError("Wrong type for pruner settings!")
return pruner
def _create_pruner(self, pruner_method: str) -> BasePruner:
if pruner_method == "halving":
pruner = SuccessiveHalvingPruner(min_resource=1,
reduction_factor=4,
min_early_stopping_rate=0)
elif pruner_method == "median":
pruner = MedianPruner(n_startup_trials=self.n_startup_trials,
n_warmup_steps=self.n_evaluations // 3)
elif pruner_method == "none":
# Do not prune
pruner = NopPruner()
else:
raise ValueError(f"Unknown pruner: {pruner_method}")
return pruner
def _create_study(mo_study: "multi_objective.study.MultiObjectiveStudy") -> "optuna.Study":
study = create_study(
storage=mo_study._storage,
sampler=_MultiObjectiveSamplerAdapter(mo_study.sampler),
pruner=NopPruner(),
study_name="_motpe-" + mo_study._storage.get_study_name_from_id(mo_study._study_id),
directions=mo_study.directions,
load_if_exists=True,
)
for mo_trial in mo_study.trials:
with warnings.catch_warnings():
warnings.simplefilter("ignore", ExperimentalWarning)
study.add_trial(_create_trial(mo_trial))
return study
def create_study(
directions: List[str],
study_name: Optional[str] = None,
storage: Optional[Union[str, BaseStorage]] = None,
sampler: Optional["multi_objective.samplers.BaseMultiObjectiveSampler"] = None,
load_if_exists: bool = False,
) -> "multi_objective.study.MultiObjectiveStudy":
"""Create a new :class:`~optuna.multi_objective.study.MultiObjectiveStudy`.
Example:
.. testcode::
import optuna
def objective(trial):
# Binh and Korn function.
x = trial.suggest_float("x", 0, 5)
y = trial.suggest_float("y", 0, 3)
v0 = 4 * x ** 2 + 4 * y ** 2
v1 = (x - 5) ** 2 + (y - 5) ** 2
return v0, v1
study = optuna.multi_objective.create_study(["minimize", "minimize"])
study.optimize(objective, n_trials=3)
Args:
directions:
Optimization direction for each objective value.
Set ``minimize`` for minimization and ``maximize`` for maximization.
study_name:
Study's name. If this argument is set to None, a unique name is generated
automatically.
storage:
Database URL. If this argument is set to None, in-memory storage is used, and the
:class:`~optuna.study.Study` will not be persistent.
.. note::
When a database URL is passed, Optuna internally uses `SQLAlchemy`_ to handle
the database. Please refer to `SQLAlchemy's document`_ for further details.
If you want to specify non-default options to `SQLAlchemy Engine`_, you can
instantiate :class:`~optuna.storages.RDBStorage` with your desired options and
pass it to the ``storage`` argument instead of a URL.
.. _SQLAlchemy: https://www.sqlalchemy.org/
.. _SQLAlchemy's document:
https://docs.sqlalchemy.org/en/latest/core/engines.html#database-urls
.. _SQLAlchemy Engine: https://docs.sqlalchemy.org/en/latest/core/engines.html
sampler:
A sampler object that implements background algorithm for value suggestion.
If :obj:`None` is specified,
:class:`~optuna.multi_objective.samplers.NSGAIIMultiObjectiveSampler` is used
as the default. See also :class:`~optuna.multi_objective.samplers`.
load_if_exists:
Flag to control the behavior to handle a conflict of study names.
In the case where a study named ``study_name`` already exists in the ``storage``,
a :class:`~optuna.exceptions.DuplicatedStudyError` is raised if ``load_if_exists`` is
set to :obj:`False`.
Otherwise, the creation of the study is skipped, and the existing one is returned.
Returns:
A :class:`~optuna.multi_objective.study.MultiObjectiveStudy` object.
"""
# TODO(ohta): Support pruner.
mo_sampler = sampler or multi_objective.samplers.NSGAIIMultiObjectiveSampler()
sampler_adapter = multi_objective.samplers._MultiObjectiveSamplerAdapter(mo_sampler)
if not isinstance(directions, Iterable):
raise TypeError("`directions` must be a list or other iterable types.")
if not all(d in ["minimize", "maximize"] for d in directions):
raise ValueError("`directions` includes unknown direction names.")
study = _create_study(
study_name=study_name,
storage=storage,
sampler=sampler_adapter,
pruner=NopPruner(),
load_if_exists=load_if_exists,
)
study.set_system_attr("multi_objective:study:directions", list(directions))
return MultiObjectiveStudy(study)
