Objects of class boost_family
define negative gradients of
loss functions to be optimized.
Objects can be created by calls of the form Family(...)
ngradient
:a function with arguments y
and f
implementing the negative gradient of
the loss
function.
risk
:a risk function with arguments y
, f
and w
,
the weighted mean of the loss function by default.
offset
:a function with argument y
and w
(weights)
for computing a scalar offset.
weights
:a logical indicating if weights are allowed.
check_y
:a function for checking the class / mode of a response variable.
nuisance
:a function for extracting nuisance parameters.
response
:inverse link function of a GLM or any other transformation on the scale of the response.
rclass
:function to derive class predictions from conditional class probabilities (for models with factor response variable).
name
:a character giving the name of the loss function for pretty printing.
charloss
:a character, the deparsed loss function.
Family