GAM logistic backend for xplus

Additive logistic models via mgcv::gam() with a smoothing spline per feature. Fractional pseudo-labels are fitted with a quasibinomial family, binary targets with the binomial family.

Usage

xplus_backend_gam(k = NULL, method = "REML", max_features = 100)

Arguments

k

Basis dimension passed to each mgcv::s() term; NULL uses the mgcv default.

method

Smoothing-parameter estimation method for binary targets (fractional targets always use "GCV.Cp", required by the quasibinomial family).

max_features

Maximum number of features accepted; GAM cost grows quickly with dimension.

Value

An "xplus_backend" object.

See also

xplus_backend(), xplus()

Examples

if (FALSE) { # \dontrun{
fit <- xplus(x, y, backend = xplus_backend_gam())
} # }