Base

systematica.portfolio.transformers.base

BaseTransformer

BaseTransformer()

Base class for transforming trading metrics.

Ancestors

• abc.ABC

Descendants

• systematica.portfolio.transformers.logistic.Logistic

Methods

transform

transform(
    self,
    metrics: Dict[str, float],
) ‑> Dict[str, float]

Transform metrics based on the selected transformation method.

Parameters:

Name	Type	Default	Description
metrics	tp.Dict[str, float]	--	A dictionary containing performance metrics, where keys are metric names and values are their corresponding values.

Returns:

Type	Description
tp.Dict[str, float]	A dictionary of transformed metrics, where keys are the same as input and values are the transformed values.

Reward

Reward(
    transformer: systematica.portfolio.transformers.base.BaseTransformer,
    weights: Dict[str, float] = _Nothing.NOTHING,
)

Computes a composite reward score.

This class computes a reward score as a weighted sum of processed metrics. It uses a processor to transform the raw metrics and applies weights to each metric. The weight for ‘max_drawdown’ is inverted because a lower value is preferable.

Method generated by attrs for class Reward.

Instance variables

• transformer: systematica.portfolio.transformers.base.BaseTransformer: An instance of a transformer to process the metrics. If None, defaults to Logistic.

• weights: Dict[str, float]: A dictionary mapping metric names to weights. Defaults to an empty dictionary.

Methods

run

run(
    self,
    metrics: Dict[str, float],
) ‑> float

Calculate the composite reward score.

This method adjusts specific metrics (e.g., inverting ‘max_drawdown’), processes the metrics using the assigned processor, and computes a weighted sum of the processed metrics to derive a composite reward score.

Parameters:

Name	Type	Default	Description
metrics	tp.Dict[str, float]	--	A dictionary containing performance metrics. If max_drawdown is present, its value will be inverted prior to processing.

Returns:

Type	Description
float	The computed composite reward score.