Log¶
Logarithmic loss.
This loss function expects each provided y_pred to be a logit. In other words if must be the raw output of a linear model or a neural network.
Parameters¶
-
weight_pos – defaults to
1.0 -
weight_neg – defaults to
1.0
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).