loss_LKL_xgb: Laurae's Kullback-Leibler Error (xgboost function)

In Laurae2/Laurae: Advanced High Performance Data Science Toolbox for R

Description

This function computes for xgboost's obj function the Laurae's Kullback-Leibler Error loss gradient and hessian per value provided preds and dtrain.

Usage

loss_LKL_xgb(preds, dtrain)

Arguments

preds	The predictions.
dtrain	The xgboost model.

Details

This loss function is strictly positive, therefore defined in \]0, +Inf\[. It penalizes lower values more heavily, and as such is a good fit for typical problems requiring fine tuning when undercommitting on the predictions. Compared to Laurae's Poisson loss function, Laurae's Kullback-Leibler loss has much higher loss. Negative and null values are set to 1e-15. This loss function is experimental.

Loss Formula : (y_true - y_pred) * log(y_true / y_pred)

Gradient Formula : -((y_true - y_pred)/y_pred + log(y_true) - log(y_pred))

Hessian Formula : ((y_true - y_pred)/y_pred + 2)/y_pred

Value

The gradient and the hessian of the Laurae's Kullback-Leibler Error per value in a list.

Laurae2/Laurae documentation built on May 8, 2019, 7:59 p.m.