boost_control: Specify Controls for the Boosting Algorithm

Description Usage Arguments Value See Also

View source: R/helpers.R

Description

Specify control and tuning parameters of the boosting algorithm.

Usage

1
2
3
boost_control(mstop = 100, nu = 0.1, maxit = 30000, risk = c("inbag","oobag", "none"),
    which.offset = c("mle", "zero"), savedata = TRUE, trace = TRUE,
    parallel = require("multicore"), hardStop = TRUE, save_ensembless=TRUE)

Arguments

mstop

integer. Initial number of boosting iterations

nu

double. Step-size between 0 and 1

maxit

integer. Controls the maximum number of iterations that are used for numerical optimazation via optim(..., method = "BFGS").

risk

character. Determines how the empirical risk should be computed. It can take the values risk = "inbag", i.e., the risk is computed for the learning sample, risk = "oobag", i.e., the risk is computed for the validation sample and risk = "none".

which.offset

character. Indictating the choice of offset. This can be either the maximum likelihood estimator of the null model (which.offset = "mle") or zero (which.offset = "zero").

savedata

logic. Indicating whether the data should be saved in the returned cfboost object.

trace

logic. Determining if status information should be printed during estimation of the model.

parallel

logic. enable parallelization using multicore technology for crossvalidation (in function cv). If multicore is available parallelization is used per default but can be switched off.

hardStop

logic. Indicates if the initial value of mstop is the maximal number of iterations (mstop = TRUE) or if it should be increased if the algorithm did not converge until mstop. The latter is only possible with risk = "oobag".

save_ensembless

logic. Indicating if the list of coefficients for the selected base-learners should be saved and returned. This list is generally needed but can be suppressed to reduce memory usage (not recommended).

Value

An object of class boost_control is returned (as a list).

See Also

cfboost


CoxFlexBoost documentation built on May 2, 2019, 6:53 p.m.