ImplicitQuantileNetworkDistributionLoss

class pytorch_forecasting.metrics.distributions.ImplicitQuantileNetworkDistributionLoss(quantiles: list[float] | None = None, input_size: int | None = 16, hidden_size: int | None = 32, n_loss_samples: int | None = 64)

Bases: DistributionLoss

Implicit Quantile Network Distribution Loss.

Based on Probabilistic Time Series Forecasting with Implicit Quantile Networks. A network is used to directly map network outputs to a quantile.

Initialize implicit quantile network distribution loss.

Parameters:

• quantiles (list[float], optional) – Default quantiles to output. Defaults to [0.02, 0.1, 0.25, 0.5, 0.75, 0.9, 0.98].

• input_size (int, optional) – Input size per prediction length. Defaults to 16.

• hidden_size (int, optional) – Hidden size per prediction length. Defaults to 32.

• n_loss_samples (int, optional) – Number of quantiles to sample to calculate loss.

Methods

loss(y_pred, y_actual)	Calculate negative likelihood.
rescale_parameters(parameters, target_scale, ...)	Rescale normalized parameters into the scale required for the output.
sample(y_pred, n_samples)	Sample from distribution.
to_prediction(y_pred[, n_samples])	Convert network prediction into a point prediction.
to_quantiles(y_pred[, quantiles])	Convert network prediction into a quantile prediction.

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

Calculate negative likelihood.

Parameters:

• y_pred (torch.Tensor) – Network output.

• y_actual (torch.Tensor) – Actual values.

Returns:

Metric value on which backpropagation can be applied.

Return type:

torch.Tensor

rescale_parameters(parameters: Tensor, target_scale: Tensor, encoder: BaseEstimator) → Tensor

Rescale normalized parameters into the scale required for the output.

Parameters:

• parameters (torch.Tensor) – normalized parameters (indexed by last dimension)

• target_scale (torch.Tensor) – scale of parameters (n_batch_samples x (center, scale))

• encoder (BaseEstimator) – original encoder that normalized the target in the first place

Returns:

parameters in real/not normalized space

Return type:

torch.Tensor

sample(y_pred, n_samples: int) → Tensor

Sample from distribution.

Parameters:

• y_pred – prediction output of network (shape batch_size x n_timesteps x n_parameters)

• n_samples (int) – number of samples to draw

Returns:

tensor with samples (shape batch_size x n_timesteps x n_samples)

Return type:

torch.Tensor

to_prediction(y_pred: Tensor, n_samples: int = 100) → Tensor

Convert network prediction into a point prediction.

Parameters:

• y_pred – prediction output of network

• n_samples (int) – number of samples to draw

Returns:

mean prediction

Return type:

torch.Tensor

to_quantiles(y_pred: Tensor, quantiles: list[float] = None) → Tensor

Convert network prediction into a quantile prediction.

Parameters:

• y_pred (torch.Tensor) – Prediction output of network.

• quantiles (list[float], optional) – Quantiles for probability range. Defaults to quantiles defined in the class initialization.

Returns:

Prediction quantiles (last dimension).

Return type:

torch.Tensor