GradientScript.jl

This script gives an example to compute loglikehood and its gradient (often called as a back projection, and proportional to the traditional residual map in CLEAN methods) from interferometric visibilities.

The script will use Comrade.jl to compute loglikelihood, and take a uvfits file to load visibilities. The user optionally sets an input image for which loglikelihood will be evaluated. Otherwise an empty image will be used for complex visibilities, or a Gaussian image with the unit total flux density will be used for closure quantities.

Julia version and setup

To use this we assume that you have Julia 1.10 installed. If you do not we recommend installing Julia with juliaup https://github.com/JuliaLang/juliaup. If you have juliaup install to use Julia 1.10 run the following commands

juliaup add lts
juliaup default lts

The main script to run is main.jl. To use it first call setup.jl:

julia setup.jl

Usage

To compute the gradient, you can run the command

julia main.jl <uvfits> -o <path-to-output> ...

It will output the gradient image in a FITS file. <uvfits> indicates the path to the input uvfits for which the likelihood and its gradient to be computed. Other options are:

• -o, --outpath: the path to the output directory where images and other stats will be saved. Default is the current directory.
• --fitsfile: the input fits image, to which the gradient will be computed.
• --fovx: the field of view in microarcseconds. Default is 200 μas.
• --fovy: the field of view in microarcseconds. Default is fovx.
• -p, --psize: the pixel size in microarcseconds. Default is 1 μas.
• -x, --x: the x offset of image center in microarcseconds. Default is 0 μas.
• -y, --y: the y offset of image center in microarcseconds. Default is 0 μas.
• -u, --uvmin: the minimum uv distance in λ. Default is 0.2e9.
• --scanavg: the average data at each scan at the loading
• --ferr: the fractional error in the data. Default is 0.0.
• --closure: use the closure quantities (lca, cp) instead of complex visibilities for the likelihood
• --logimage: use the logintensity for the image parameters
• --comshift: include the center-of-mass shift in the sky model