import copy
from typing import Any
from typing import Dict
from typing import List
from typing import Optional
from typing import Tuple
import numpy as np
import optuna
from optuna._experimental import experimental
from optuna._imports import try_import
from optuna import distributions
from optuna import samplers
from optuna.samplers import BaseSampler
from optuna.study import Study
from optuna.study import StudyDirection
from optuna.trial import FrozenTrial
from optuna.trial import TrialState
with try_import() as _imports:
import skopt
from skopt.space import space
[docs]class SkoptSampler(BaseSampler):
"""Sampler using Scikit-Optimize as the backend.
Example:
Optimize a simple quadratic function by using :class:`~optuna.integration.SkoptSampler`.
.. testcode::
import optuna
def objective(trial):
x = trial.suggest_uniform('x', -10, 10)
y = trial.suggest_int('y', 0, 10)
return x**2 + y
sampler = optuna.integration.SkoptSampler()
study = optuna.create_study(sampler=sampler)
study.optimize(objective, n_trials=10)
Args:
independent_sampler:
A :class:`~optuna.samplers.BaseSampler` instance that is used for independent
sampling. The parameters not contained in the relative search space are sampled
by this sampler.
The search space for :class:`~optuna.integration.SkoptSampler` is determined by
:func:`~optuna.samplers.intersection_search_space()`.
If :obj:`None` is specified, :class:`~optuna.samplers.RandomSampler` is used
as the default.
.. seealso::
:class:`optuna.samplers` module provides built-in independent samplers
such as :class:`~optuna.samplers.RandomSampler` and
:class:`~optuna.samplers.TPESampler`.
warn_independent_sampling:
If this is :obj:`True`, a warning message is emitted when
the value of a parameter is sampled by using an independent sampler.
Note that the parameters of the first trial in a study are always sampled
via an independent sampler, so no warning messages are emitted in this case.
skopt_kwargs:
Keyword arguments passed to the constructor of
`skopt.Optimizer <https://scikit-optimize.github.io/#skopt.Optimizer>`_
class.
Note that ``dimensions`` argument in ``skopt_kwargs`` will be ignored
because it is added by :class:`~optuna.integration.SkoptSampler` automatically.
n_startup_trials:
The independent sampling is used until the given number of trials finish in the
same study.
consider_pruned_trials:
If this is :obj:`True`, the PRUNED trials are considered for sampling.
.. note::
Added in v2.0.0 as an experimental feature. The interface may change in newer
versions without prior notice. See
https://github.com/optuna/optuna/releases/tag/v2.0.0.
.. note::
As the number of trials :math:`n` increases, each sampling takes longer and longer
on a scale of :math:`O(n^3)`. And, if this is :obj:`True`, the number of trials
will increase. So, it is suggested to set this flag :obj:`False` when each
evaluation of the objective function is relatively faster than each sampling. On
the other hand, it is suggested to set this flag :obj:`True` when each evaluation
of the objective function is relatively slower than each sampling.
"""
[docs] def __init__(
self,
independent_sampler: Optional[BaseSampler] = None,
warn_independent_sampling: bool = True,
skopt_kwargs: Optional[Dict[str, Any]] = None,
n_startup_trials: int = 1,
*,
consider_pruned_trials: bool = False
) -> None:
_imports.check()
self._skopt_kwargs = skopt_kwargs or {}
if "dimensions" in self._skopt_kwargs:
del self._skopt_kwargs["dimensions"]
self._independent_sampler = independent_sampler or samplers.RandomSampler()
self._warn_independent_sampling = warn_independent_sampling
self._n_startup_trials = n_startup_trials
self._search_space = samplers.IntersectionSearchSpace()
self._consider_pruned_trials = consider_pruned_trials
if self._consider_pruned_trials:
self._raise_experimental_warning_for_consider_pruned_trials()
@experimental("2.0.0", name="`consider_pruned_trials = True` in SkoptSampler")
def _raise_experimental_warning_for_consider_pruned_trials(self) -> None:
pass
[docs] def reseed_rng(self) -> None:
self._independent_sampler.reseed_rng()
[docs] def infer_relative_search_space(
self, study: Study, trial: FrozenTrial
) -> Dict[str, distributions.BaseDistribution]:
search_space = {}
for name, distribution in self._search_space.calculate(study).items():
if distribution.single():
if not isinstance(distribution, distributions.CategoricalDistribution):
# `skopt` cannot handle non-categorical distributions that contain just
# a single value, so we skip this distribution.
#
# Note that `Trial` takes care of this distribution during suggestion.
continue
search_space[name] = distribution
return search_space
[docs] def sample_relative(
self,
study: Study,
trial: FrozenTrial,
search_space: Dict[str, distributions.BaseDistribution],
) -> Dict[str, Any]:
if len(search_space) == 0:
return {}
complete_trials = self._get_trials(study)
if len(complete_trials) < self._n_startup_trials:
return {}
optimizer = _Optimizer(search_space, self._skopt_kwargs)
optimizer.tell(study, complete_trials)
return optimizer.ask()
[docs] def sample_independent(
self,
study: Study,
trial: FrozenTrial,
param_name: str,
param_distribution: distributions.BaseDistribution,
) -> Any:
if self._warn_independent_sampling:
complete_trials = self._get_trials(study)
if len(complete_trials) >= self._n_startup_trials:
self._log_independent_sampling(trial, param_name)
return self._independent_sampler.sample_independent(
study, trial, param_name, param_distribution
)
def _log_independent_sampling(self, trial: FrozenTrial, param_name: str) -> None:
logger = optuna.logging.get_logger(__name__)
logger.warning(
"The parameter '{}' in trial#{} is sampled independently "
"by using `{}` instead of `SkoptSampler` "
"(optimization performance may be degraded). "
"You can suppress this warning by setting `warn_independent_sampling` "
"to `False` in the constructor of `SkoptSampler`, "
"if this independent sampling is intended behavior.".format(
param_name, trial.number, self._independent_sampler.__class__.__name__
)
)
def _get_trials(self, study: Study) -> List[FrozenTrial]:
complete_trials = []
for t in study.get_trials(deepcopy=False):
if t.state == TrialState.COMPLETE:
complete_trials.append(t)
elif (
t.state == TrialState.PRUNED
and len(t.intermediate_values) > 0
and self._consider_pruned_trials
):
_, value = max(t.intermediate_values.items())
if value is None:
continue
# We rewrite the value of the trial `t` for sampling, so we need a deepcopy.
copied_t = copy.deepcopy(t)
copied_t.value = value
complete_trials.append(copied_t)
return complete_trials
class _Optimizer(object):
def __init__(
self, search_space: Dict[str, distributions.BaseDistribution], skopt_kwargs: Dict[str, Any]
) -> None:
self._search_space = search_space
dimensions = []
for name, distribution in sorted(self._search_space.items()):
if isinstance(distribution, distributions.UniformDistribution):
# Convert the upper bound from exclusive (optuna) to inclusive (skopt).
high = np.nextafter(distribution.high, float("-inf"))
dimension = space.Real(distribution.low, high)
elif isinstance(distribution, distributions.LogUniformDistribution):
# Convert the upper bound from exclusive (optuna) to inclusive (skopt).
high = np.nextafter(distribution.high, float("-inf"))
dimension = space.Real(distribution.low, high, prior="log-uniform")
elif isinstance(distribution, distributions.IntUniformDistribution):
count = (distribution.high - distribution.low) // distribution.step
dimension = space.Integer(0, count)
elif isinstance(distribution, distributions.IntLogUniformDistribution):
low = distribution.low - 0.5
high = distribution.high + 0.5
dimension = space.Real(low, high, prior="log-uniform")
elif isinstance(distribution, distributions.DiscreteUniformDistribution):
count = int((distribution.high - distribution.low) // distribution.q)
dimension = space.Integer(0, count)
elif isinstance(distribution, distributions.CategoricalDistribution):
dimension = space.Categorical(distribution.choices)
else:
raise NotImplementedError(
"The distribution {} is not implemented.".format(distribution)
)
dimensions.append(dimension)
self._optimizer = skopt.Optimizer(dimensions, **skopt_kwargs)
def tell(self, study: Study, complete_trials: List[FrozenTrial]) -> None:
xs = []
ys = []
for trial in complete_trials:
if not self._is_compatible(trial):
continue
x, y = self._complete_trial_to_skopt_observation(study, trial)
xs.append(x)
ys.append(y)
self._optimizer.tell(xs, ys)
def ask(self) -> Dict[str, Any]:
params = {}
param_values = self._optimizer.ask()
for (name, distribution), value in zip(sorted(self._search_space.items()), param_values):
if isinstance(distribution, distributions.DiscreteUniformDistribution):
value = value * distribution.q + distribution.low
if isinstance(distribution, distributions.IntUniformDistribution):
value = value * distribution.step + distribution.low
if isinstance(distribution, distributions.IntLogUniformDistribution):
value = int(np.round(value))
value = min(max(value, distribution.low), distribution.high)
params[name] = value
return params
def _is_compatible(self, trial: FrozenTrial) -> bool:
# Thanks to `intersection_search_space()` function, in sequential optimization,
# the parameters of complete trials are always compatible with the search space.
#
# However, in distributed optimization, incompatible trials may complete on a worker
# just after an intersection search space is calculated on another worker.
for name, distribution in self._search_space.items():
if name not in trial.params:
return False
distributions.check_distribution_compatibility(distribution, trial.distributions[name])
param_value = trial.params[name]
param_internal_value = distribution.to_internal_repr(param_value)
if not distribution._contains(param_internal_value):
return False
return True
def _complete_trial_to_skopt_observation(
self, study: Study, trial: FrozenTrial
) -> Tuple[List[Any], float]:
param_values = []
for name, distribution in sorted(self._search_space.items()):
param_value = trial.params[name]
if isinstance(distribution, distributions.DiscreteUniformDistribution):
param_value = (param_value - distribution.low) // distribution.q
if isinstance(distribution, distributions.IntUniformDistribution):
param_value = (param_value - distribution.low) // distribution.step
param_values.append(param_value)
value = trial.value
assert value is not None
if study.direction == StudyDirection.MAXIMIZE:
value = -value
return param_values, value