neuraxle.metaopt.tpe

Tree parzen estimator

Code for tree parzen estimator auto ml.

Classes

TreeParzenEstimatorHyperparameterSelectionStrategy(…)

class neuraxle.metaopt.tpe.TreeParzenEstimatorHyperparameterSelectionStrategy(number_of_initial_random_step: int = 40, quantile_threshold: float = 0.3, number_good_trials_max_cap: int = 25, number_possible_hyperparams_candidates: int = 100, prior_weight: float = 0.0, use_linear_forgetting_weights: bool = False, number_recent_trial_at_full_weights: int = 25)[source]
_abc_impl = <_abc_data object>
_adaptive_parzen_normal(hyperparam_distribution, distribution_trials)[source]

This code is enterily inspire from Hyperopt (https://github.com/hyperopt) code.

_create_gaussian_mixture(continuous_distribution: neuraxle.hyperparams.distributions.HyperparameterDistribution, trial_hyperparameters: List[neuraxle.hyperparams.space.HyperparameterSamples])[source]
_create_posterior(flat_hyperparameter_space: neuraxle.hyperparams.space.HyperparameterSpace, trials: neuraxle.metaopt.trial.Trials) → neuraxle.hyperparams.space.HyperparameterSpace[source]
_reweights_categorical(discrete_distribution: neuraxle.hyperparams.distributions.DiscreteHyperparameterDistribution, trial_hyperparameters)[source]
find_next_best_hyperparams(auto_ml_container: neuraxle.metaopt.auto_ml.AutoMLContainer) → neuraxle.hyperparams.space.HyperparameterSamples[source]

Find the next best hyperparams using previous trials.

Parameters

auto_ml_container (Trials) – trials data container

Returns

next best hyperparams

Return type

HyperparameterSamples

neuraxle.metaopt.tpe._linear_forgetting_Weights(number_samples, number_recent_trial_at_full_weights)[source]

This code has been taken from Hyperopt (https://github.com/hyperopt) code.