Control function for xrnet
fitting.
xrnet.control( tolerance = 1e-08, max_iterations = 1e+05, dfmax = NULL, pmax = NULL, lower_limits = NULL, upper_limits = NULL )
tolerance | positive convergence criterion. Default is 1e-08. |
---|---|
max_iterations | maximum number of iterations to run coordinate gradient descent across all penalties before returning an error. Default is 1e+05. |
dfmax | maximum number of variables allowed in model. Default is \(ncol(x) + ncol(unpen) + ncol(external) + intercept[1] + intercept[2]\). |
pmax | maximum number of variables with nonzero coefficient estimate. Default is \(min(2 * dfmax + 20, ncol(x) + ncol(unpen) + ncol(external) + intercept[2])\). |
lower_limits | vector of lower limits for each coefficient. Default is -Inf for all variables. |
upper_limits | vector of upper limits for each coefficient. Default is Inf for all variables. |
A list object with the following components:
The coordinate descent stopping criterion.
The maximum number of variables that will be allowed in the model.
The maximum number of variables with nonzero coefficient estimate.
Feature-specific numeric vector of lower bounds for coefficient estimates
Feature-specific numeric vector of upper bounds for coefficient estimates