multiJGL: Linear and nonlinear multiclass network estimation with joint...

In KontioJuho/multiJGL: What the Package Does (One Line, Title Case)

Linear and nonlinear multiclass network estimation with joint regularization over categorical groups

Description

Linear and nonlinear multiclass network estimation with joint regularization over categorical groups

Usage

multiJGL(
  node.covariates = node.covariates,
  grouping.factor = grouping.factor,
  penalty.lin = "fused",
  penalty.nonlin = "fused",
  lin_lambda1 = 0.1,
  lin_lambda2 = 0.025,
  nonlin_lambda1 = 0.1,
  nonlin_lambda2 = 0.025,
  tol.linear = 1e-05,
  tol.nonlinear = 1e-05,
  subsample.pseudo_obs = FALSE,
  omit.rate = 2L,
  ...
)

Arguments

node.covariates	An nxp dimensional matrix of p covariates measured over n samples.
grouping.factor	A grouping factor for creating observational classes.
penalty.lin	Specify "fused" or "group" penalty type for the linear JGL algorithm.
penalty.nonlin	Specify "fused" or "group" penalty type for the nonlinear JGL algorithm. #See the explanation for the fused and group penalties in the JGL package #The original JGL CRAN repository: https://CRAN.R-project.org/package=JGL #The following penalty parameters are given in pairs to – #separately assign the amount of regularizations for linear and nonlinear parts
lin_lambda1	The l1-penalty parameter for the linear JGL to regulate within group network densities
lin_lambda2	The l1-penalty parameter for the nonlinear JGL.
nonlin_lambda1	The fusion penalty parameter for the linear JGL.
nonlin_lambda2	The fusion penalty parameter for the nonlinear JGL.
tol.linear	Convergence criterion for the linear part (see the JGL package for details).
tol.nonlinear	Convergence criterion for the nonlinear part. #Subsampling procedure over pseudo-observations if the number of observations is large already in the original sets.
subsample.pseudo_obs	Should the subsampling procedure be used over the pseudo-observations.
omit.rate	An integer: Omit rate for the subsampling pcocedure between 2L and 5L
...	Additional parameter for the nonlinear JGL.

Examples

 print("net <- multiJGL(node.covariates, grouping.factor)")

KontioJuho/multiJGL documentation built on Oct. 30, 2022, 3:42 a.m.