youcef4k / muffin.jl Goto Github PK

A new version of Muffin.jl running on a cluster is under heavy development and will be available soon. As a foretaste, see the paper submitted to EUSIPCO 2016 conference : http://arxiv.org/abs/1602.08847.

=========

Muffin.jl is a julia implementation of MUFFIN: Multi-Frequency Sparse Radio Interferometric imaging

see http://arxiv.org/abs/1504.06847

#Getting Started

####Installation

Muffin.jl uses the following Julia packages :

• FITSIO.jl : reading and writing FITS.
• Images.jl : Fast Fourier Transform implementation.
• Wavelets.jl : package for fast wavelet transforms.
• HDF5.jl : for writing JLD ("Julia data") variables.
• GHF.jl : generation of Gaussian homogeneous spatial field.
• PyPlot.jl : provides a Julia interface to the Matplotlib plotting library.

The installation is as simple as :

 Pkg.clone("https://github.com/andferrari/GHF.jl.git")  
 Pkg.add("PyPlot")   
 Pkg.add("FITSIO")   
 Pkg.add("Images")    
 Pkg.add("Wavelets")   
 Pkg.add("HDF5")    

To install Muffin.jl, type from a Julia session the following command :

	Pkg.clone("https://github.com/andferrari/Muffin.jl.git")

####Usage

To load the Muffin.jl module, type from a Julia session :

	using Muffin

To use parallel computing, start Julia with nprocs local process and load the module :

	$ julia -p nprocs
	julia> @everywhere using Muffin

You just need to add the keyword parallel in the MUFFIN function, for an example :

	julia> psfst, skyst, algost, admmst, toolst = muffin(nitermax = 10,parallel="true");

#Functions

####Main Function muffin(...)

• muffin is called with parameters definition :

	  psfst, skyst, algost, admmst, toolst = muffin(folder, dataobj, datapsf, nitermax, rhop, rhot, rhov, rhos, μt, μv, mueps)

• It returns 5 structures :

  • psfst : datas related to the PSF
  • skyst : datas related to the SKY
  • algost : contains algorithm parameters
  • admmst : datas related to the admm method
  • toolst : contains error calculation arrays

• Parameters are :

  • folder : path to the folder containing FITS files. Default : installation folder, containing FITS files for the demo
  • dataobj : FITS object.
  • datapsf : FITS psf.
  • nitermax : maximum number of ADMM iterations. Default : 500
  • rhop : ADMM parameter for positivity constraint. Default : 1
  • rhot : ADMM parameter for spatial constraint. Default : 5
  • rhov : ADMM parameter for spectral constraint. Default : 2
  • rhos : ADMM parameter for spectral constraint. Default : 1
  • μt : Spatial regularization parameter. Default : 5e-1
  • μv : Spectral regularization parameter. Default : 1
  • mueps : Ridge/Tikhonov regularization parameter : 1e-3
  • ws : warm start
  • parallel : to start parallel Muffin

####Save data savedata(...)

• savedata is called by :

	savedata(savepath, psfst, skyst, algost, admmst, toolst)

where savepath is the path/JLD file name where datas will be saved.

#Muffin Demo

Muffin.jl contains a demo file muffindemo.jl in the default installation folder.

To run the demo, type :

using Muffin
demo = string(Pkg.dir("Muffin"),"/src/muffindemo.jl")
include(demo);

If you want to visualize your results, just type after the simulation :

plot = string(Pkg.dir("Muffin"),"/src/readdata.jl")
include(plot);

The demo will run with the following parameters :

myfolder = string(Pkg.dir("Muffin"))
mydataobj =  "data/M31.fits"
mydatapsf =  "data/meerkat_m30_25pix.psf.fits"
mynitermax = 1
myrhop = 1
myrhot = 5
myrhov = 2
myrhos = 1
myμt = 5e-1
myμv = 1
mymueps = 1e-3
mysavepath = string(myfolder,myfolder[1],"data/demo_results.jld")

and the results will be saved in the demo_results.jld file.