Source code for optuna.samplers.tpe.sampler

import math

import numpy as np
import scipy.special
from scipy.stats import truncnorm

from optuna import distributions
from optuna.samplers import base
from optuna.samplers import random
from optuna.samplers.tpe.parzen_estimator import _ParzenEstimator
from optuna.samplers.tpe.parzen_estimator import _ParzenEstimatorParameters
from optuna.study import StudyDirection
from optuna.trial import TrialState
from optuna import type_checking

if type_checking.TYPE_CHECKING:
    from typing import Any  # NOQA
    from typing import Callable  # NOQA
    from typing import Dict  # NOQA
    from typing import List  # NOQA
    from typing import Optional  # NOQA
    from typing import Tuple  # NOQA

    from optuna.distributions import BaseDistribution  # NOQA
    from optuna.study import Study  # NOQA
    from optuna.trial import FrozenTrial  # NOQA

EPS = 1e-12


def default_gamma(x):
    # type: (int) -> int

    return min(int(np.ceil(0.1 * x)), 25)


def hyperopt_default_gamma(x):
    # type: (int) -> int

    return min(int(np.ceil(0.25 * np.sqrt(x))), 25)


def default_weights(x):
    # type: (int) -> np.ndarray

    if x == 0:
        return np.asarray([])
    elif x < 25:
        return np.ones(x)
    else:
        ramp = np.linspace(1.0 / x, 1.0, num=x - 25)
        flat = np.ones(25)
        return np.concatenate([ramp, flat], axis=0)


[docs]class TPESampler(base.BaseSampler): """Sampler using TPE (Tree-structured Parzen Estimator) algorithm. This sampler is based on *independent sampling*. See also :class:`~optuna.samplers.BaseSampler` for more details of 'independent sampling'. On each trial, for each parameter, TPE fits one Gaussian Mixture Model (GMM) ``l(x)`` to the set of parameter values associated with the best objective values, and another GMM ``g(x)`` to the remaining parameter values. It chooses the parameter value ``x`` that maximizes the ratio ``l(x)/g(x)``. For further information about TPE algorithm, please refer to the following papers: - `Algorithms for Hyper-Parameter Optimization <https://papers.nips.cc/paper/4443-algorithms-for-hyper-parameter-optimization.pdf>`_ - `Making a Science of Model Search: Hyperparameter Optimization in Hundreds of Dimensions for Vision Architectures <http://proceedings.mlr.press/v28/bergstra13.pdf>`_ Example: .. testcode:: import optuna from optuna.samplers import TPESampler def objective(trial): x = trial.suggest_uniform('x', -10, 10) return x**2 study = optuna.create_study(sampler=TPESampler()) study.optimize(objective, n_trials=10) Args: consider_prior: Enhance the stability of Parzen estimator by imposing a Gaussian prior when :obj:`True`. The prior is only effective if the sampling distribution is either :class:`~optuna.distributions.UniformDistribution`, :class:`~optuna.distributions.DiscreteUniformDistribution`, :class:`~optuna.distributions.LogUniformDistribution`, :class:`~optuna.distributions.IntUniformDistribution`, or :class:`~optuna.distributions.IntLogUniformDistribution`. prior_weight: The weight of the prior. This argument is used in :class:`~optuna.distributions.UniformDistribution`, :class:`~optuna.distributions.DiscreteUniformDistribution`, :class:`~optuna.distributions.LogUniformDistribution`, :class:`~optuna.distributions.IntUniformDistribution`, :class:`~optuna.distributions.IntLogUniformDistribution`, and :class:`~optuna.distributions.CategoricalDistribution`. consider_magic_clip: Enable a heuristic to limit the smallest variances of Gaussians used in the Parzen estimator. consider_endpoints: Take endpoints of domains into account when calculating variances of Gaussians in Parzen estimator. See the original paper for details on the heuristics to calculate the variances. n_startup_trials: The random sampling is used instead of the TPE algorithm until the given number of trials finish in the same study. n_ei_candidates: Number of candidate samples used to calculate the expected improvement. gamma: A function that takes the number of finished trials and returns the number of trials to form a density function for samples with low grains. See the original paper for more details. weights: A function that takes the number of finished trials and returns a weight for them. See `Making a Science of Model Search: Hyperparameter Optimization in Hundreds of Dimensions for Vision Architectures <http://proceedings.mlr.press/v28/bergstra13.pdf>`_ for more details. seed: Seed for random number generator. """ def __init__( self, consider_prior=True, # type: bool prior_weight=1.0, # type: float consider_magic_clip=True, # type: bool consider_endpoints=False, # type: bool n_startup_trials=10, # type: int n_ei_candidates=24, # type: int gamma=default_gamma, # type: Callable[[int], int] weights=default_weights, # type: Callable[[int], np.ndarray] seed=None, # type: Optional[int] ): # type: (...) -> None self._parzen_estimator_parameters = _ParzenEstimatorParameters( consider_prior, prior_weight, consider_magic_clip, consider_endpoints, weights ) self._prior_weight = prior_weight self._n_startup_trials = n_startup_trials self._n_ei_candidates = n_ei_candidates self._gamma = gamma self._weights = weights self._rng = np.random.RandomState(seed) self._random_sampler = random.RandomSampler(seed=seed)
[docs] def reseed_rng(self) -> None: self._rng = np.random.RandomState() self._random_sampler.reseed_rng()
def infer_relative_search_space(self, study, trial): # type: (Study, FrozenTrial) -> Dict[str, BaseDistribution] return {} def sample_relative(self, study, trial, search_space): # type: (Study, FrozenTrial, Dict[str, BaseDistribution]) -> Dict[str, Any] return {} def sample_independent(self, study, trial, param_name, param_distribution): # type: (Study, FrozenTrial, str, BaseDistribution) -> Any values, scores = _get_observation_pairs(study, param_name, trial) n = len(values) if n < self._n_startup_trials: return self._random_sampler.sample_independent( study, trial, param_name, param_distribution ) below_param_values, above_param_values = self._split_observation_pairs(values, scores) if isinstance(param_distribution, distributions.UniformDistribution): return self._sample_uniform(param_distribution, below_param_values, above_param_values) elif isinstance(param_distribution, distributions.LogUniformDistribution): return self._sample_loguniform( param_distribution, below_param_values, above_param_values ) elif isinstance(param_distribution, distributions.DiscreteUniformDistribution): return self._sample_discrete_uniform( param_distribution, below_param_values, above_param_values ) elif isinstance(param_distribution, distributions.IntUniformDistribution): return self._sample_int(param_distribution, below_param_values, above_param_values) elif isinstance(param_distribution, distributions.IntLogUniformDistribution): return self._sample_int_loguniform( param_distribution, below_param_values, above_param_values ) elif isinstance(param_distribution, distributions.CategoricalDistribution): index = self._sample_categorical_index( param_distribution, below_param_values, above_param_values ) return param_distribution.choices[index] else: distribution_list = [ distributions.UniformDistribution.__name__, distributions.LogUniformDistribution.__name__, distributions.DiscreteUniformDistribution.__name__, distributions.IntUniformDistribution.__name__, distributions.IntLogUniformDistribution.__name__, distributions.CategoricalDistribution.__name__, ] raise NotImplementedError( "The distribution {} is not implemented. " "The parameter distribution should be one of the {}".format( param_distribution, distribution_list ) ) def _split_observation_pairs( self, config_vals, # type: List[Optional[float]] loss_vals, # type: List[Tuple[float, float]] ): # type: (...) -> Tuple[np.ndarray, np.ndarray] config_vals = np.asarray(config_vals) loss_vals = np.asarray(loss_vals, dtype=[("step", float), ("score", float)]) n_below = self._gamma(len(config_vals)) loss_ascending = np.argsort(loss_vals) below = config_vals[np.sort(loss_ascending[:n_below])] below = np.asarray([v for v in below if v is not None], dtype=float) above = config_vals[np.sort(loss_ascending[n_below:])] above = np.asarray([v for v in above if v is not None], dtype=float) return below, above def _sample_uniform(self, distribution, below, above): # type: (distributions.UniformDistribution, np.ndarray, np.ndarray) -> float low = distribution.low high = distribution.high return self._sample_numerical(low, high, below, above) def _sample_loguniform(self, distribution, below, above): # type: (distributions.LogUniformDistribution, np.ndarray, np.ndarray) -> float low = distribution.low high = distribution.high return self._sample_numerical(low, high, below, above, is_log=True) def _sample_discrete_uniform(self, distribution, below, above): # type:(distributions.DiscreteUniformDistribution, np.ndarray, np.ndarray) -> float q = distribution.q r = distribution.high - 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 # Shift below and above to [0, r] above -= distribution.low below -= distribution.low best_sample = self._sample_numerical(low, high, below, above, q=q) + distribution.low return min(max(best_sample, distribution.low), distribution.high) def _sample_int(self, distribution, below, above): # type: (distributions.IntUniformDistribution, np.ndarray, np.ndarray) -> int d = distributions.DiscreteUniformDistribution( low=distribution.low, high=distribution.high, q=distribution.step ) return int(self._sample_discrete_uniform(d, below, above)) def _sample_int_loguniform(self, distribution, below, above): # type: (distributions.IntLogUniformDistribution, np.ndarray, np.ndarray) -> int low = distribution.low - 0.5 high = distribution.high + 0.5 sample = self._sample_numerical(low, high, below, above, is_log=True) best_sample = ( np.round((sample - distribution.low) / distribution.step) * distribution.step + distribution.low ) return int(min(max(best_sample, distribution.low), distribution.high)) def _sample_numerical( self, low, # type: float high, # type: float below, # type: np.ndarray above, # type: np.ndarray q=None, # type: Optional[float] is_log=False, # type: bool ): # type: (...) -> float if is_log: low = np.log(low) high = np.log(high) below = np.log(below) above = np.log(above) size = (self._n_ei_candidates,) parzen_estimator_below = _ParzenEstimator( mus=below, low=low, high=high, parameters=self._parzen_estimator_parameters ) samples_below = self._sample_from_gmm( parzen_estimator=parzen_estimator_below, low=low, high=high, q=q, size=size, ) log_likelihoods_below = self._gmm_log_pdf( samples=samples_below, parzen_estimator=parzen_estimator_below, low=low, high=high, q=q, ) parzen_estimator_above = _ParzenEstimator( mus=above, low=low, high=high, parameters=self._parzen_estimator_parameters ) log_likelihoods_above = self._gmm_log_pdf( samples=samples_below, parzen_estimator=parzen_estimator_above, low=low, high=high, q=q, ) ret = float( TPESampler._compare( samples=samples_below, log_l=log_likelihoods_below, log_g=log_likelihoods_above )[0] ) return math.exp(ret) if is_log else ret def _sample_categorical_index(self, distribution, below, above): # type: (distributions.CategoricalDistribution, np.ndarray, np.ndarray) -> int choices = distribution.choices below = list(map(int, below)) above = list(map(int, above)) upper = len(choices) size = (self._n_ei_candidates,) weights_below = self._weights(len(below)) counts_below = np.bincount(below, minlength=upper, weights=weights_below) weighted_below = counts_below + self._prior_weight weighted_below /= weighted_below.sum() samples_below = self._sample_from_categorical_dist(weighted_below, size) log_likelihoods_below = TPESampler._categorical_log_pdf(samples_below, weighted_below) weights_above = self._weights(len(above)) counts_above = np.bincount(above, minlength=upper, weights=weights_above) weighted_above = counts_above + self._prior_weight weighted_above /= weighted_above.sum() log_likelihoods_above = TPESampler._categorical_log_pdf(samples_below, weighted_above) return int( TPESampler._compare( samples=samples_below, log_l=log_likelihoods_below, log_g=log_likelihoods_above )[0] ) def _sample_from_gmm( self, parzen_estimator, # type: _ParzenEstimator low, # type: float high, # type: float q=None, # type: Optional[float] size=(), # type: Tuple ): # type: (...) -> np.ndarray weights = parzen_estimator.weights mus = parzen_estimator.mus sigmas = parzen_estimator.sigmas weights, mus, sigmas = map(np.asarray, (weights, mus, sigmas)) if low >= high: raise ValueError( "The 'low' should be lower than the 'high'. " "But (low, high) = ({}, {}).".format(low, high) ) active = np.argmax(self._rng.multinomial(1, weights, size=size), axis=-1) trunc_low = (low - mus[active]) / sigmas[active] trunc_high = (high - mus[active]) / sigmas[active] while True: samples = truncnorm.rvs( trunc_low, trunc_high, size=size, loc=mus[active], scale=sigmas[active], random_state=self._rng, ) if (samples < high).all(): break if q is None: return samples else: return np.round(samples / q) * q def _gmm_log_pdf( self, samples, # type: np.ndarray parzen_estimator, # type: _ParzenEstimator low, # type: float high, # type: float q=None, # type: Optional[float] ): # type: (...) -> np.ndarray weights = parzen_estimator.weights mus = parzen_estimator.mus sigmas = parzen_estimator.sigmas samples, weights, mus, sigmas = map(np.asarray, (samples, weights, mus, sigmas)) if samples.size == 0: return np.asarray([], dtype=float) if weights.ndim != 1: raise ValueError( "The 'weights' should be 2-dimension. " "But weights.shape = {}".format(weights.shape) ) if mus.ndim != 1: raise ValueError( "The 'mus' should be 2-dimension. " "But mus.shape = {}".format(mus.shape) ) if sigmas.ndim != 1: raise ValueError( "The 'sigmas' should be 2-dimension. " "But sigmas.shape = {}".format(sigmas.shape) ) p_accept = np.sum( weights * ( TPESampler._normal_cdf(high, mus, sigmas) - TPESampler._normal_cdf(low, mus, sigmas) ) ) if q is None: distance = samples[..., None] - mus mahalanobis = (distance / np.maximum(sigmas, EPS)) ** 2 Z = np.sqrt(2 * np.pi) * sigmas coefficient = weights / Z / p_accept return TPESampler._logsum_rows(-0.5 * mahalanobis + np.log(coefficient)) else: cdf_func = TPESampler._normal_cdf upper_bound = np.minimum(samples + q / 2.0, high) lower_bound = np.maximum(samples - q / 2.0, low) probabilities = np.sum( weights[..., None] * ( cdf_func(upper_bound[None], mus[..., None], sigmas[..., None]) - cdf_func(lower_bound[None], mus[..., None], sigmas[..., None]) ), axis=0, ) return np.log(probabilities + EPS) - np.log(p_accept + EPS) def _sample_from_categorical_dist(self, probabilities, size): # type: (np.ndarray, Tuple[int]) -> np.ndarray if probabilities.size == 1 and isinstance(probabilities[0], np.ndarray): probabilities = probabilities[0] probabilities = np.asarray(probabilities) if size == (0,): return np.asarray([], dtype=float) assert len(size) assert probabilities.ndim == 1 n_draws = int(np.prod(size)) sample = self._rng.multinomial(n=1, pvals=probabilities, size=int(n_draws)) assert sample.shape == size + (probabilities.size,) return_val = np.dot(sample, np.arange(probabilities.size)) return_val.shape = size return return_val @classmethod def _categorical_log_pdf( cls, sample, # type: np.ndarray p, # type: np.ndarray ): # type: (...) -> np.ndarray if sample.size: return np.log(np.asarray(p)[sample]) else: return np.asarray([]) @classmethod def _compare(cls, samples, log_l, log_g): # type: (np.ndarray, np.ndarray, np.ndarray) -> np.ndarray samples, log_l, log_g = map(np.asarray, (samples, log_l, log_g)) if samples.size: score = log_l - log_g if samples.size != score.size: raise ValueError( "The size of the 'samples' and that of the 'score' " "should be same. " "But (samples.size, score.size) = ({}, {})".format(samples.size, score.size) ) best = np.argmax(score) return np.asarray([samples[best]] * samples.size) else: return np.asarray([]) @classmethod def _logsum_rows(cls, x): # type: (np.ndarray) -> np.ndarray x = np.asarray(x) m = x.max(axis=1) return np.log(np.exp(x - m[:, None]).sum(axis=1)) + m @classmethod def _normal_cdf(cls, x, mu, sigma): # type: (float, np.ndarray, np.ndarray) -> np.ndarray mu, sigma = map(np.asarray, (mu, sigma)) denominator = x - mu numerator = np.maximum(np.sqrt(2) * sigma, EPS) z = denominator / numerator return 0.5 * (1 + scipy.special.erf(z)) @classmethod def _log_normal_cdf(cls, x, mu, sigma): # type: (float, np.ndarray, np.ndarray) -> np.ndarray mu, sigma = map(np.asarray, (mu, sigma)) if x < 0: raise ValueError("Negative argument is given to _lognormal_cdf. x: {}".format(x)) denominator = np.log(np.maximum(x, EPS)) - mu numerator = np.maximum(np.sqrt(2) * sigma, EPS) z = denominator / numerator return 0.5 + 0.5 * scipy.special.erf(z)
[docs] @staticmethod def hyperopt_parameters(): # type: () -> Dict[str, Any] """Return the the default parameters of hyperopt (v0.1.2). :class:`~optuna.samplers.TPESampler` can be instantiated with the parameters returned by this method. Example: Create a :class:`~optuna.samplers.TPESampler` instance with the default parameters of `hyperopt <https://github.com/hyperopt/hyperopt/tree/0.1.2>`_. .. testcode:: import optuna from optuna.samplers import TPESampler def objective(trial): x = trial.suggest_uniform('x', -10, 10) return x**2 sampler = TPESampler(**TPESampler.hyperopt_parameters()) study = optuna.create_study(sampler=sampler) study.optimize(objective, n_trials=10) Returns: A dictionary containing the default parameters of hyperopt. """ return { "consider_prior": True, "prior_weight": 1.0, "consider_magic_clip": True, "consider_endpoints": False, "n_startup_trials": 20, "n_ei_candidates": 24, "gamma": hyperopt_default_gamma, "weights": default_weights, }
def _get_observation_pairs(study, param_name, trial): # type: (Study, str, FrozenTrial) -> Tuple[List[Optional[float]], List[Tuple[float, float]]] """Get observation pairs from the study. This function collects observation pairs from the complete or pruned trials of the study. The values for trials that don't contain the parameter named ``param_name`` are set to None. An observation pair fundamentally consists of a parameter value and an objective value. However, due to the pruning mechanism of Optuna, final objective values are not always available. Therefore, this function uses intermediate values in addition to the final ones, and reports the value with its step count as ``(-step, value)``. Consequently, the structure of the observation pair is as follows: ``(param_value, (-step, value))``. The second element of an observation pair is used to rank observations in ``_split_observation_pairs`` method (i.e., observations are sorted lexicographically by ``(-step, value)``). """ sign = 1 if study.direction == StudyDirection.MAXIMIZE: sign = -1 values = [] scores = [] for trial in study.get_trials(deepcopy=False): if trial.state is TrialState.COMPLETE and trial.value is not None: score = (-float("inf"), sign * trial.value) elif trial.state is TrialState.PRUNED: if len(trial.intermediate_values) > 0: step, intermediate_value = max(trial.intermediate_values.items()) if math.isnan(intermediate_value): score = (-step, float("inf")) else: score = (-step, sign * intermediate_value) else: score = (float("inf"), 0.0) else: continue param_value = None # type: Optional[float] if param_name in trial.params: distribution = trial.distributions[param_name] param_value = distribution.to_internal_repr(trial.params[param_name]) values.append(param_value) scores.append(score) return values, scores