Find_ridge_best_split: Find_ridge_best_split

In theo-s/Rforestry_R: Random Forests

Description

This is the fast impelementation of the best split function when the leaf prediciton is coming from ridge regression.

Usage

Find_ridge_best_split(feat, y, linear.idx, current.splitting.idx, lambda,
  update_A_inv = update_A_inv_R)

Arguments

feat	all features
y	outcome vector
linear.idx	feature indexes where a Ridge Regression should be fitted
current.splitting.idx	the index of the feature we are currently testing for splits.

Value

It returns a vector of two doubles which determines what the best splitting point is, and what it's MSE is.

split_val_best	the best splitting value
MSE_best	the best MSE

Examples

set.seed(309814)
n <- 1000
feat <- matrix(rnorm(4 * n), ncol = 4)
feat[2,] <- feat[which.min(feat[,2]),]
feat[3,] <- feat[which.min(feat[,2]),]
feat[4,] <- feat[which.min(feat[,2]),]
feat[5,] <- feat[which.min(feat[,2]),] - 1
y <- rnorm(n)
linear.idx <- 1:4
current.splitting.idx <- 2
lambda <- .2
Find_ridge_best_split_linear_MASS(feat = feat, y = y, linear.idx = linear.idx,
                                 current.splitting.idx = current.splitting.idx,
                                 lambda = lambda)
Find_ridge_best_split_slow(feat = feat, y = y, linear.idx = linear.idx,
                          current.splitting.idx = current.splitting.idx,
                          lambda = lambda)
Find_ridge_best_split(feat = feat, y = y, linear.idx = linear.idx,
                     current.splitting.idx = current.splitting.idx,
                     lambda = lambda)
# Note that fast split is not estimating the RSS or the MSE, but the RSS -
# sum(y^2).

theo-s/Rforestry_R documentation built on Dec. 23, 2021, 9:55 a.m.