hhcart_reflect_feature_space_g: hhcart_reflect_feature_space_g

In hhcartr: HHCART(G) - A Reflected Feature Space for CART

Description

This function is in an internal only function. It is used to reflect the feature data if the data is found to be suitable, otherwise node splitting will be done using the original data and using axis-parallel splits.

Usage

hhcart_reflect_feature_space_g(
  X,
  y,
  useIdentity = FALSE,
  max_features = "None",
  n_features = n_features,
  n_classes = n_classes,
  classify = TRUE,
  depth = depth,
  colx = 1
)

Arguments

X	The feature variables of the current node being processed.
y	The target variable for the corresponding feature variables.
useIdentity	A flag, if FALSE the node split will be done using an oblique split after transforming the training data using a householder transform. If TRUE the 'transform' will be the identity and the split will be done on the original data.
max_features	The maximum number of features to consider in the current split.
n_features	The number of feature variables.
n_classes	The number of classes in the target variable.
classify	A flag, if TRUE this is a classification problem, if FALSE this is a regression problem.
depth	The current tree depth.
colx	The eigenvector column to use.

Details

The following parameters are supported:

X_matrix, X, y, most_freq_class, n_classes, max_features, n_features, colx

Value

Returns a list containing idx, thr, X_house, householder_matrix, using_householder if split was for a classification problem and reflected data used, otherwise the results from split_using_original_data are returned if useIdentity was TRUE. For a regression problem a list containing the following is returned: idx, thr, as.matrix(X), NULL, FALSE.

hhcartr documentation built on July 2, 2021, 9:06 a.m.