Quantile¶
Quantile loss.
Parameters¶
-
alpha – defaults to
0.5Desired quantile to attain.
Examples¶
>>> from river import optim
>>> loss = optim.losses.Quantile(0.5)
>>> loss(1, 3)
1.0
>>> loss.gradient(1, 3)
0.5
>>> loss.gradient(3, 1)
-0.5
Methods¶
call
Returns the loss.
Parameters
- y_true
- y_pred
Returns
The loss(es).
gradient
Return the gradient with respect to y_pred.
Parameters
- y_true
- y_pred
Returns
The gradient(s).
mean_func
Mean function.
This is the inverse of the link function. Typically, a loss function takes as input the raw output of a model. In the case of classification, the raw output would be logits. The mean function can be used to convert the raw output into a value that makes sense to the user, such as a probability.
Parameters
- y_pred
Returns
The adjusted prediction(s).