Pluggable losses for the StemGNN forecaster.

Point losses keep the shared signature (pred, target) -> scalar and reduce by mean. MultiQuantileLoss is the multi-quantile (pinball) loss. Every loss carries an outputsize_multiplier so the network's output head width is loss-driven. All are module-level / class-based so a fitted estimator pickles cleanly.

mae

stemgnn_losses.mae

Mean absolute error.

Parameter	Type	Default	Description
pred	jnp.ndarray	-	(undocumented)
target	jnp.ndarray	-	(undocumented)

Returns: jnp.ndarray (scalar mean loss).

mse

stemgnn_losses.mse

Mean squared error.

Parameter	Type	Default	Description
pred	jnp.ndarray	-	(undocumented)
target	jnp.ndarray	-	(undocumented)

Returns: jnp.ndarray (scalar mean loss).

huber

stemgnn_losses.huber

Huber loss, delta = 1.0.

Parameter	Type	Default	Description
pred	jnp.ndarray	-	(undocumented)
target	jnp.ndarray	-	(undocumented)

Returns: jnp.ndarray (scalar mean loss).

MultiQuantileLoss

stemgnn_losses.MultiQuantileLoss

Multi-quantile (pinball) loss. __call__(pred[...,h,Q], target[...,h]).

QL(y, y_hat, q) = q*(y-y_hat)+ + (1-q)*(y_hat-y)+, averaged over quantiles and all elements. Quantiles are sorted, must lie in (0, 1), and must include 0.5 (the median / "mean" head). Picklable (holds a plain tuple).

__init__(self, quantiles=(0.1, 0.5, 0.9))

Parameter	Type	Default	Description
quantiles	Sequence[float]	(0.1, 0.5, 0.9)	(undocumented)

__call__(self, pred, target) -> jnp.ndarray

Parameter	Type	Default	Description
pred	jnp.ndarray	-	(undocumented)
target	jnp.ndarray	-	(undocumented)

Returns: jnp.ndarray (scalar mean loss).

outputsize_multiplier

stemgnn_losses.outputsize_multiplier

Output-head width implied by loss (1 for point losses).

Parameter	Type	Default	Description
loss	-	-	(undocumented)

Returns: int.

resolve

stemgnn_losses.resolve

Return a callable loss from a registry string, a callable, or an MQ instance.

Parameter	Type	Default	Description
loss	str | LossFn | MultiQuantileLoss	-	(undocumented)

Returns: Callable loss function.