Fit symmetric LS-SVR with RMSPE loss (deprecated)

Soft-deprecated wrapper retained for backwards compatibility. New code should use psvr(loss = "rmspe", sym = +1L, ...) (or sym = -1L). Returns the legacy psvr_rmspe_sym object shape.

Usage

rmspe_sym_lssvr(X, y, kernel, gamma, a = 1, precondition = "auto")

Arguments

X, y: Training matrix and strictly-positive target vector.
kernel: Kernel closure from make_kernel().
gamma: Regularization parameter Γ > 0.
a: Symmetry parameter: 1 (even) or -1 (odd).
precondition: One of "auto", "always", "never", or a positive numeric threshold; controls Remark-17 symmetric rescaling.

Value

A list of class "psvr_rmspe_sym".

Examples

if (FALSE) { # \dontrun{
fit <- psvr(X, y, loss = "rmspe", sym = +1L,
            kernel = make_kernel("rbf"), gamma = 100)
} # }