xtune.control {xtune}

Control function for xtune fitting

Description

Control function for xtune fitting.

Usage

xtune.control(
  alpha.est.init = NULL,
  max_s = 20,
  margin_s = 1e-05,
  maxstep = 100,
  margin = 0.001,
  maxstep_inner = 100,
  margin_inner = 0.001,
  compute.likelihood = FALSE,
  verbosity = FALSE,
  standardize = TRUE,
  intercept = TRUE
)

Arguments

alpha.est.init

Initial values of alpha vector supplied to the algorithm. Alpha values are the hyper-parameters for the double exponential prior of regression coefficients, and it controls the prior variance of regression coefficients. Default is a vector of 0 with length p.

max_s

Maximum number of outer loop iterations for binary or multiclass outcomes. Default is 20.

margin_s

Convergence threshold of the outer loop for binary or multiclass outcomes. Default is 1e-5.

maxstep

Maximum number of iterations. Default is 100.

margin

Convergence threshold. Default is 1e-3.

maxstep_inner

Maximum number of iterations for the inner loop of the majorization-minimization algorithm. Default is 100.

margin_inner

Convergence threshold for the inner loop of the majorization-minimization algorithm. Default is 1e-3.

compute.likelihood

Should the function compute the marginal likelihood for hyper-parameters at each step of the update? Default is TRUE.

verbosity

Track algorithm update process? Default is FALSE.

standardize

Standardize X or not, same as the standardized option in glmnet.

intercept

Should intercept(s) be fitted (default=TRUE) or set to zero (FALSE), same as the intercept option in glmnet.

Value

A list of control objects after the checking.