Skip to contents

Candidate values

Source: R/group_lasso.R
set_candidates.Rd

Sets candidate values for hyperparameters.

Usage

.set_candidates(tune)

Arguments

tune

character string for determining the candidate values for the hyperparameters:

  • "none": fixed weights and exponents (wgt_local=1, exp_local=0, wgt_global=0), no tuning (share information within groups, ignore correlations)

  • "weight" (default): fixed exponents (exp_local=0, exp_global=2), tuning wgt_local=1-wgt_global (find a compromise between sharing information within groups and between correlated predictors)

  • "exponent": fixed weights (wgt_local=1, wgt_global=0), tuning exp_local (share information between correlated predictors in the same group, determine level of trust in correlation coefficients)

  • "bivariate": tuning wgt_local=1-wgt_global and exp_local=exp_global (find a compromise between sharing information between predictors in the same group and between all predictors, determine level of trust in correlation coefficients)

  • "factorial": tuning wgt_local=1-wgt_global, exp_local, and exp_global (find a compromise between sharing information within groups and between correlated predictors, with possibly different exponents)

(The internal function .set_candidates() uses this argument to create a grid of candidates values for wgt_local, exp_local, wgt_global, and exp_global.)

Value

Returns a data frame with the slots "wgt_local" and "exp_local" for the local prior information and the slots "wgt_global" and "exp_global" for the global prior information.

Details

  • If the local weight equals 0, the local exponent has no influence. And if the global weight equals 0, the global exponent has no influence. Therefore, if a weight is close to zero, its exponent is set to infinity. This avoids redundant combinations of hyperparameters (i.e., a local weight of zero with multiple local exponents, or a global weight of zero with multiple global exponents)

  • The experimental hyperparameter "threshold" is currently always set to 0 (no thresholding of correlation coefficients).

See also

This function is called by cv.corila().

Examples

.set_candidates(tune = "none")
#>   wgt_local exp_local wgt_global exp_global threshold
#> 1         1         0          0        Inf         0