fusionlab.nn.losses.combined_total_loss¶

fusionlab.nn.losses.combined_total_loss(quantiles, anomaly_layer, anomaly_scores)[source]¶

Create a total loss that adds quantile loss + anomaly loss.

Like above, only this top-level is decorated. The returned function is not re-decorated.

Parameters:

quantiles (list of float) – Quantiles for the quantile loss part.
anomaly_layer (tf.keras.losses.Loss) – A custom Loss or callable implementing anomaly loss.
anomaly_scores (tf.Tensor or np.ndarray) – The anomaly scores needed for the anomaly loss.

Returns:

A loss function: loss_fn(y_true, y_pred)

Return type:

callable