Skip to content

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).

References