Objective

class emflow.problems.objective.Objective

Bases: ABC

abstract calculate()

Subclasses must implement this method.

class emflow.problems.objective.MeanSquaredError

Bases: Objective

property name

calculate(y_true, y_pred, mean=True, squared=True)

Mean squared error (or RMSE when squared=False), nan-aware.

NaN entries in either y_true or y_pred are ignored pairwise. Returns a scalar when mean=True, else the element-wise errors.

class emflow.problems.objective.MeanAbsoluteError

Bases: Objective

property name

calculate(y_true, y_pred, mean=True)

Mean absolute error, nan-aware (NaNs ignored pairwise).

class emflow.problems.objective.PinballLoss(quantiles)

Bases: Objective

property name

calculate(y_true, y_preds, mean=True)

Compute the pinball loss between true values and multiple sets of predictions. Each set of predictions corresponds to a specific quantile. :param y_true: array-like, true values. :param y_preds: 2D array-like, predicted values for each quantile. Shape: (n_samples, n_quantiles). :return: numpy array, the pinball losses for each quantile.