- Notifications
You must be signed in to change notification settings - Fork2
Monte Carlo Penalty Selection for graphical lasso
License
NotificationsYou must be signed in to change notification settings
markkukuismin/MCPeSe
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
Monte Carlo Penalty Selection (MCPeSe) for graphical lasso.
library(huge)source("mcpese.R")########################################################### Initialize seed number:#seed = Sys.time()#seed = as.integer(seed)#seed = seed %% 100000seed=59040set.seed(seed)########################################################### Graphical model simulation:p=200n=210Model="hub"HugeData= huge.generator(n=n,d=p,graph=Model)nlambda=100########################################################### Compute Glasso solution path:L= huge(HugeData$data,nlambda=nlambda,method="glasso")########################################################### Run the A-R selection (uniform prior),ARSelect= mcpese(L,n=n,M=1000)names(ARSelect)rhos=ARSelect$rhosARSelect$accept.rate########################################################### Run the M-H selection (uniform prior),MHSelect= mcpese(L,n=n,nSteps=1000,method="M-H")names(MHSelect)rhos=MHSelect$rhosMHSelect$accept.rateplot(rhos,type="l")##########################################################
# Either use the mean of rho values...mean(rhos)#ThetaARSelect = huge(Y, lambda=mean(rhos), method="glasso")#ThetaAR = as.matrix(ThetaARSelect$icov[[1]])# ... or pick the smallest tuning parameter value from the solution path# which is larger or equal to the mean value:optARrhoIndx=ARSelect$opt.index# This is sup{i : rho[i] >= mean(rhos)}optMHrhoIndx=MHSelect$opt.indexhuge.plot(HugeData$theta)title("Ground truth")##########################################################
huge.plot(L$path[[optARrhoIndx]])title("Accept-Reject sampling")##########################################################
huge.plot(L$path[[optMHrhoIndx]])title("Metropolis-Hastings sampling")
The MCPeSe method is described in:
Kuismin and Sillanpaa (2020). MCPeSe: Monte Carlo penalty selection for graphical lasso,Bioinformatics,https://doi.org/10.1093/bioinformatics/btaa734.
File "CodeCollection.zip" is a collection of scripts used to prepare the material in this paper.
About
Monte Carlo Penalty Selection for graphical lasso
Topics
Resources
License
Uh oh!
There was an error while loading.Please reload this page.
Stars
Watchers
Forks
Releases
No releases published
Packages0
No packages published