JointNets (version 2.0.1)

kdiffnet: Fast and Scalable Estimator for Using Additional Knowledge in Learning Sparse Structure Change of High Dimensional of Sparse Changes in High-Dimensional Gaussian Graphical Models

Description

The kdiffnet algorithm

Usage

kdiffnet(C, D, W, g = rep(1, 100), epsilon = 1, lambda = 0.05,
  knowledgeType = "EV", gamma = 4, covType = "cov",
  intertwined = FALSE, thre = "soft", rho = 0.05, iterMax = 20)

Arguments

C

A input matrix for the 'control' group. It can be data matrix or covariance matrix. If C is a symmetric matrix, the matrices are assumed to be covariance matrix.

D

A input matrix for the 'disease' group. It can be data matrix or covariance matrix. If D is a symmetric matrix, the matrices are assumed to be covariance matrix.

W

known edge level additional knowledge. It is a square matrix of dimension p X p where p is the input dimension.

g

known node level additional knowledge. It is a vector of dimension 1 X p where p is the input dimension, each entry indicating membership of node to a group, 0 for a node belonging to no group. For example, in a dataset with dimension=3,g=c(0,1,1) indicates node 1 belongs to no group, and node 2 and node 3 belong to group index 1.

epsilon

A positive number. The hyperparameter controls the sparsity level of the groups in g of the difference matrix

lambda

A positive number. The hyperparameter controls the sparsity level of the difference matrix

knowledgeType

"EV": if use overlapping node and edge level additional knowledge,"E": if only edge level additional knowledge or "V": only group level knowledge

gamma

: A positive number. This hyperparameter is used in calculating each proximity during optimization

covType

A parameter to decide which Graphical model we choose to estimate from the input data.

If covType = "cov", it means that we estimate multiple sparse Gaussian Graphical models. This option assumes that we calculate (when input X represents data directly) or use (when X elements are symmetric representing covariance matrices) the sample covariance matrices as input to the simule algorithm.

If covType = "kendall", it means that we estimate multiple nonparanormal Graphical models. This option assumes that we calculate (when input X represents data directly) or use (when X elements are symmetric representing correlation matrices) the kendall's tau correlation matrices as input to the simule algorithm.

intertwined

indicate whether to use intertwined covariance matrix

thre

A parameter to decide which threshold function to use for \(T_v\). If thre = "soft", it means that we choose soft-threshold function as \(T_v\). If thre = "hard", it means that we choose hard-threshold function as \(T_v\).

rho

A positive number. This hyperparameter controls the learning rate of the proximal gradient method.

iterMax

An integer. The max number of iterations in the optimization of the proximal algorithm

Value

$graphs

A matrix of the estimated sparse changes between two Gaussian Graphical Models

$share

null

Examples

# NOT RUN {
library(JointNets)
data(exampleData)
result = kdiffnet(exampleData[[1]], exampleData[[2]],
W = matrix(1,20,20), g = rep(0,20),epsilon = 0.2,
lambda = 0.4,covType = "cov")
plot(result)
# }