L2Boost: Implementation of the L2Boosting algorithm

In newboost: New boosting algorithms

Description

This function implements a classical version of the L2Boosting algorithm.

The L2Boost algorithm is started at different values and then the intersection of all selected variables from each run is included in the final model.

Usage

L2Boost(X, y, iter = 200, beta.start = rep(0, dim(X)[2]), post = FALSE)

L2Boost.multistart(X, y, num.start = 5, iter = 200)

Arguments

X	matrix of regressors
y	dependent variable
iter	number of iterations
beta.start	initial value for the algorithm to start with
post	logical, if post Boosting algorithm should be applied (default FALSE).

Value

The functions returns an object of class L2Boost with the following components:

• BetaFinalthe estimated final parameter vector.

• BetaVectorvector where each component gives the values of the estimated coefficient in each round.

• Resmatrix with the estimated function values in each round

• Res_betamatrix with the estimated coefficient vectors from each round

• iIteration when the process was terminated.

• Svector of the indices of the selected variables in each round

• sigma2estimation of the variance

• stop_ruleWhen the stopping rule, stopped the algorithm.

• itermaximal number of iterations of the algorithm

• ymodel response / dependent variable

• Xmodel matrix

newboost documentation built on May 2, 2019, 5:17 p.m.