Differential Network Estimation in R.

dineR

The goal of dineR is to enable users of all backgrounds to easily and computationally efficiently perform differential network estimation.

Installation

You can install the released version of dineR from CRAN with:

install.packages("dineR")

Example

This is a basic example which shows you how to solve a common problem:

library(dineR)
# Data Generation
n_X <- 100
n_Y <- n_X
p_X <- 100
p_Y <- p_X
#case <- "sparse"
case <- "asymsparse"

data <- data_generator(n = n_X, p = p_X, seed = 123)

X <- data$X
Y <- data$Y
diff_Omega <- data$diff_Omega
paste("The number of non-zero entries in the differential network is: ", sum(diff_Omega!=0))

# Estimation Preliminaries (All of the parameters are now optional as the function has pre-specified defaults)

loss <- "lasso"
nlambda <- 50
tuning <- "AIC"
stop_tol <- 1e-4
perturb <- F
correlation <- F
max_iter <- 500
lambda_min_ratio <- 0.5
#gamma <- 1 #Only if we use EBIC

# Estimation

result <- estimation(X, Y, loss = loss, nlambda = nlambda, tuning = tuning, stop_tol = stop_tol,
                      perturb = perturb, correlation = correlation,
                      max_iter = max_iter, lambda_min_ratio = lambda_min_ratio) 

# Results

print(result$path[[1]][1:5,1:5])
result$elapse