import numpy
from optuna import distributions
from optuna.samplers import BaseSampler
from optuna import type_checking
if type_checking.TYPE_CHECKING:
from typing import Any # NOQA
from typing import Dict # NOQA
from typing import Optional # NOQA
from optuna.distributions import BaseDistribution # NOQA
from optuna.study import Study # NOQA
from optuna.trial import FrozenTrial # NOQA
[docs]class RandomSampler(BaseSampler):
"""Sampler using random sampling.
This sampler is based on *independent sampling*.
See also :class:`~optuna.samplers.BaseSampler` for more details of 'independent sampling'.
Example:
.. testcode::
import optuna
from optuna.samplers import RandomSampler
def objective(trial):
x = trial.suggest_uniform('x', -5, 5)
return x**2
study = optuna.create_study(sampler=RandomSampler())
study.optimize(objective, n_trials=10)
Args:
seed: Seed for random number generator.
"""
[docs] def __init__(self, seed=None):
# type: (Optional[int]) -> None
self._rng = numpy.random.RandomState(seed)
[docs] def reseed_rng(self) -> None:
self._rng = numpy.random.RandomState()
[docs] def infer_relative_search_space(self, study, trial):
# type: (Study, FrozenTrial) -> Dict[str, BaseDistribution]
return {}
[docs] def sample_relative(self, study, trial, search_space):
# type: (Study, FrozenTrial, Dict[str, BaseDistribution]) -> Dict[str, Any]
return {}
[docs] def sample_independent(self, study, trial, param_name, param_distribution):
# type: (Study, FrozenTrial, str, distributions.BaseDistribution) -> Any
if isinstance(param_distribution, distributions.UniformDistribution):
return self._rng.uniform(param_distribution.low, param_distribution.high)
elif isinstance(param_distribution, distributions.LogUniformDistribution):
log_low = numpy.log(param_distribution.low)
log_high = numpy.log(param_distribution.high)
return float(numpy.exp(self._rng.uniform(log_low, log_high)))
elif isinstance(param_distribution, distributions.DiscreteUniformDistribution):
q = param_distribution.q
r = param_distribution.high - param_distribution.low
# [low, high] is shifted to [0, r] to align sampled values at regular intervals.
low = 0 - 0.5 * q
high = r + 0.5 * q
s = self._rng.uniform(low, high)
v = numpy.round(s / q) * q + param_distribution.low
# v may slightly exceed range due to round-off errors.
return float(min(max(v, param_distribution.low), param_distribution.high))
elif isinstance(param_distribution, distributions.IntUniformDistribution):
# [low, high] is shifted to [0, r] to align sampled values at regular intervals.
r = (param_distribution.high - param_distribution.low) / param_distribution.step
# numpy.random.randint includes low but excludes high.
s = self._rng.randint(0, r + 1)
v = s * param_distribution.step + param_distribution.low
return int(v)
elif isinstance(param_distribution, distributions.IntLogUniformDistribution):
log_low = numpy.log(param_distribution.low - 0.5)
log_high = numpy.log(param_distribution.high + 0.5)
s = numpy.exp(self._rng.uniform(log_low, log_high))
v = numpy.round(s)
return int(min(max(v, param_distribution.low), param_distribution.high))
elif isinstance(param_distribution, distributions.CategoricalDistribution):
choices = param_distribution.choices
index = self._rng.randint(0, len(choices))
return choices[index]
else:
raise NotImplementedError