QuantileLoss

class pytorch_forecasting.metrics.quantile.QuantileLoss(quantiles: List[float] = [0.02, 0.1, 0.25, 0.5, 0.75, 0.9, 0.98], **kwargs)

Bases: MultiHorizonMetric

Quantile loss, i.e. a quantile of q=0.5 will give half of the mean absolute error as it is calculated as

Defined as max(q * (y-y_pred), (1-q) * (y_pred-y))

Quantile loss

Parameters:

quantiles – quantiles for metric

Methods

loss(y_pred, target)	Calculate loss without reduction.
to_prediction(y_pred)	Convert network prediction into a point prediction.
to_quantiles(y_pred)	Convert network prediction into a quantile prediction.

loss(y_pred: Tensor, target: Tensor) → Tensor

Calculate loss without reduction. Override in derived classes

Parameters:

• y_pred – network output

• y_actual – actual values

Returns:

loss/metric as a single number for backpropagation

Return type:

torch.Tensor

to_prediction(y_pred: Tensor) → Tensor

Convert network prediction into a point prediction.

Parameters:

y_pred – prediction output of network

Returns:

point prediction

Return type:

torch.Tensor

to_quantiles(y_pred: Tensor) → Tensor

Convert network prediction into a quantile prediction.

Parameters:

y_pred – prediction output of network

Returns:

prediction quantiles

Return type:

torch.Tensor