cv.KLR: Cross-validation on Kernel Logistic Regression
In fontaine618/KLR: Kernel Logistic Regression

Description Usage Arguments Details Value See Also

View source: R/cv.R

This function performs cross-validation to select values of tuning parameters.

cv.KLR(
  y,
  x,
  n_folds = 5,
  kernel = c("gaussian", "polynomial")[1],
  lambda = c(1, 0.1, 0.01, 0.001),
  sigma2 = c(5, 2, 1, 0.5),
  d = 3,
  threshold = 1e-06,
  max_iter = 1e+05
)

`y`	A `n x 1` column vector containing the responses (0-1).
`x`	A `n x p` matrix containing the covariates.
`n_folds`	Number of folds in the CV.
`kernel`	The kernel to use. Either `gaussian` (default) or `polynomial`.
`lambda`	The regularization parameter(s).
`sigma2`	The scale(s) in the `gaussian` and `polynomial` kernel. See details.
`d`	The degree(s) in the `polynomial` kernel.
`threshold`	The convergence threshold.
`max_iter`	The maximum number of iterations.

The gaussian kernel has the following form:

exp(-||x-y||^2/sigma2).

The polynomial kernel has the following form:

(1+x'y/sigma2)^d.

A list containing:

mpe: The mean prediction error across all folds for all combinations of parameters.
lambda_min: The selected value of lambda.
sigma2_min: The selected value of sigma2.
d_min: The selected value of d.