Optuna Tuner

systematica.portfolio.tuners.optuna_tuner

OptunaTuner

OptunaTuner(
    pruner: optuna.pruners._base.BasePruner = <optuna.pruners._successive_halving.SuccessiveHalvingPruner object>,
    sampler: optuna.samplers._base.BaseSampler = <optuna.samplers._tpe.sampler.TPESampler object>,
    direction: str | List[str] = 'maximize',
    create_study_kwargs: Dict[str, Any] = _Nothing.NOTHING,
    n_trials: int = 100,
    n_completed_trials: int = None,
    n_jobs: int = -1,
    gc_after_trial: bool = False,
    optimize_kwargs: Dict[str, Any] = _Nothing.NOTHING,
    verbose: bool = True,
    tracker: systematica.portfolio.trackers.base.BaseTracker = None,
    metrics: str | List[str] = 'sharpe_ratio',
    metric_registry_kwargs: Dict[str, Any] = _Nothing.NOTHING,
    cross_validate: bool = False,
    reduce_func: Callable = CPUDispatcher(<function nanmedian_nb>),
    use_rolling: bool = False,
)

Method generated by attrs for class OptunaTuner.

Static methods

get_all_metrics

get_all_metrics(
    use_rolling: bool = False,
) ‑> List[str]

Get a list of metric keys from the registry that have a valid rolling_func. Optionally include non-tunable metrics based on user choice.

Parameters:

Name	Type	Default	Description
use_rolling	bool	False	Flag to use rolling method via BaseRollingReport. Note that both rolling and cv models can be passed. Defaults to False.

Returns:

Type	Description
List[str]	List of metric keys with valid (non-NaN) rolling_func values.

Instance variables

• target_names: List[str]: Normalize metric names.

• create_study_kwargs: Dict[str, Any]: Additional optuna study parameters.

• cross_validate: bool: Whether to use cross-validation analytics through run_report, wrapping BaseReportCV or BaseRollingReport instead of using run_pipeline alone. If True, optimize using cross validation. Otherwise, uses data history as a whole. Defaults to False.

• direction: str | List[str]: Optimization direction for Optuna (default: ‘maximize’). If list, direction is modified to “directions” internally.

• gc_after_trial: bool: Flag to determine whether to automatically run garbage collection after each trial. Set to True to run the garbage collection, False otherwise. It runs a full collection by internally calling gc.collect. If you see an increase in memory consumption over several trials, try setting this flag to True.

• metric_registry_kwargs: Dict[str, Any]: Additional arguments for metric registry. Parameters such as window of minp for rolling metrics should be specified at this level.

• metrics: str | List[str]: Performance metrics to compute (default: ‘sharpe_ratio’).

• n_completed_trials: int: Number of trials to complete before stopping (default: None).

• n_jobs: int: Parallel jobs for Optuna (default: -1).

• n_trials: int: Number of optimization trials for Optuna (default: 100).

• optimize_kwargs: Dict[str, Any]: Additional arguments for Optuna optimization.

• pruner: optuna.pruners._base.BasePruner: Pruner. Defaults to Asynchronous Successive Halving Algorithm.

• reduce_func: Callable: Reduce function when using cross_validate. If None, defaults to vbt.nb.nanmedian_nb. Defaults to None.

• sampler: optuna.samplers._base.BaseSampler: Sampler. Defaults to TPE (Tree-structured Parzen Estimator) algorithm.

• tracker: systematica.portfolio.trackers.base.BaseTracker: Log and monitor the Optuna hyperparameter using tracker integration.

• use_rolling: bool: Flag to use rolling method via BaseRollingReport. Note that both rolling and cv models can be passed. Defaults to False.

• verbose: bool: Enable verbose output (default: True).

Methods

check_direction

check_direction(
    self,
    attribute: str,
    value: str | list[str],
)

Validate optimization directions.

Parameters:

Name	Type	Default	Description
attribute	attrs.Attribute	--	The attribute being validated.
value	BaseStatArb	--	The model instance.

Raises:

Type	Description
ValueError	If direction is not “maximize” or “minimize”.
TypeError	If direction is not a string or a list

check_metrics

check_metrics(
    self,
    attribute: str,
    value: str | list[str],
)

Validate metrics

Parameters:

Name	Type	Default	Description
attribute	attrs.Attribute	--	The attribute being validated.
value	BaseStatArb	--	The model instance.

Raises:

Type	Description
TypeError	If metrics is not a string or list.
ValueError	If metric has not been registred and is not tunable.

create_objective

create_objective(
    self,
    feature_config: FeatureConfig,
) ‑> Callable

Creates an optimized objective function for Optuna by separating fixed and tunable parameters once.

Parameters:

Name	Type	Default	Description
feature_config	FeatureConfig	--	Feature configurations.

Returns:

Type	Description
tp.Callable	A callable objective function for Optuna optimization.

run_feature

run_feature(
    self,
    feature_config: FeatureConfig,
) ‑> optuna.study.study.Study

Run feature config.

Parameters:

Name	Type	Default	Description
feature_config	FeatureConfig	--	feature configuration.

Returns:

Type	Description
optuna.Study	Optuna study.

run

run(
    self,
    objective: Callable,
) ‑> optuna.study.study.Study

Run optuna study or tracker.

Parameters:

Name	Type	Default	Description
objective	tp.Callable	--	Objective function.

Returns:

Type	Description
study : optuna.Study	Optuna Study.

run_tracker

run_tracker(
    self,
    objective: Callable,
) ‑> optuna.study.study.Study

Run tracker.

Parameters:

Name	Type	Default	Description
objective	tp.Callable	--	Objective function.

Returns:

Type	Description
study : optuna.Study	Optuna Study.

run_optuna

run_optuna(
    self,
    objective: Callable,
) ‑> optuna.study.study.Study

Run optuna.

Parameters:

Name	Type	Default	Description
objective	tp.Callable	--	Objective function.

Returns:

Type	Description
study : optuna.Study	Optuna Study.