Resampling-based conditional Gaussian Bayesian Network
Usage
ensemble_cgbn(
dat,
discrete_variable = NULL,
num_iteration = 1,
boot = FALSE,
sub_ratio = 0.9,
sample_class = NULL,
hugin = FALSE,
constraints = NULL,
alpha = 0.01,
tol = 1e-04,
maxit = 0,
n_cores = NULL
)Arguments
- dat
a n x p data frame/matrix with row and column names, n = number of samples, and p = number of features
- discrete_variable
the columns that represents the discrete_variable, must be a character vector
- num_iteration
an integer, indicating the number of iteration of resampling
- boot
a logical variable, if TRUE, then perform bootstrap resampling, else, perform subsampling
- sub_ratio
a numerical value between 0 and 1, indicating the subsampling ratio
- sample_class
a atomic vector, indicating the class of each sample, if != NULL, then stratified sampling is performed
- hugin
a logical variable, if TRUE, the Hugin object will be included in the output
- constraints
prior structure knowledge, default = NULL, details -> RHugin::learn.structure
- alpha
parameter of PC algorithm for structure learning
- tol
parameter of EM algorithm for CPT learning, details -> RHugin::learn.cpt
- maxit
parameter of EM algorithm for CPT learning, details -> RHugin::learn.cpt
- n_cores
number of cores for parallel computing