Hyperparameter Tuning API¶
doubleml_pyspark.utils._tune_optuna.DMLOptunaResult
dataclass
¶
Container for Optuna search results.
This dataclass holds the results of Optuna-based hyperparameter tuning, including the best estimator, parameters, score, and the complete study history.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
learner_name
|
str
|
Name of the learner passed (e.g., 'ml_g'). |
required |
best_estimator
|
object
|
The estimator instance with the best found hyperparameters set and fitted on the full dataset used during tuning. |
required |
tuned
|
bool
|
Indicates whether tuning was performed (True) or skipped (False). |
required |
best_params
|
dict
|
The best hyperparameters found during tuning. Empty dict for untuned models. |
None
|
best_score
|
float | None
|
The best average cross-validation score achieved during tuning. None for untuned models. |
None
|
study
|
Study | None
|
The Optuna study object containing the tuning history. None for untuned models. |
None
|
scoring_method
|
str or callable
|
The scoring method used during tuning. None for untuned models. |
None
|
doubleml_pyspark.utils._tune_optuna._dml_tune_optuna
¶
_dml_tune_optuna(spark_session: SparkSession, estimator: Any, param_space_fn: Callable[[Trial], dict[str, Any]], train_df_features: DataFrame, test_df_features: DataFrame, label_col: str, learner_name: str, scoring_methods: dict, optuna_settings: dict | None = None, storage_level: StorageLevel = StorageLevel.MEMORY_AND_DISK, use_persist: bool = False) -> DMLOptunaResult
Perform Optuna hyperparameter tuning for a given estimator using PySpark DataFrames.