some-lisa-tools-julia

A very small collection of LISA-relevant Julia codes. Includes contributions from Maria Jose Bustamante Rosell (@majoburo) and Robbie Rosati (@rjrosati)

Setting up

First install Julia, then clone this repo. Open a terminal in the cloned repo's folder. Start Julia with several threads (for the demo to work, we'll want 4).

$ julia -t 4

As a small aside, the Julia workflow is a little different than python -- because Julia uses just-in-time compilation, you'll avoid excessive recompilation if you leave a julia REPL running between editing and rerunning code.

Activating the environment

Next we'll want to activate the environment in this folder

julia> import Pkg; Pkg.activate(".")

(as a shortcut, you can also press the ] key and type activate .).

Instantiating the environment

The first time you run this, you'll also need install the dependencies and precompile them (you only have to do this once):

julia> Pkg.instantiate()

This will take a little while.

Then you'll be able to run the codes here.

How to run

Each time you re-open Julia, you'll need to reactivate the environment in this folder before running any of the codes.

There are currently two mostly unrelated things here:

• A parallel-tempered MCMC code in pt.jl
• A few utilities for loading and plotting the sangria data in sangria.jl

Parallel-tempered MCMC

This code is a mostly textbook implementation of a parallel-tempered MCMC. It is based off of a geophysics review paper.

You can run a test of the code (implementing a fit of the data from Hogg et al (2010)) with

julia> include("pt.jl")
julia> demo()

This will create the file corner_pt.png with a corner plot of the samples.

Some caveats which could affect how you want to use the code, but are fixable:

• the number of threads and the number of chains are currently assumed to be identical
• the priors are currently assumed to be a product distribution (i.e. all variables' priors are independent)
• the proposals are currently hardcoded to a truncated gaussian distribution (stdev in each axis equal to the prior's stdev / 10 )

Sangria reading code

This code just reads the Sangria training data (assumed to have been downloaded to the current folder), processes it a bit and makes plots.

The data loaded into memory and then converted to the A,E,T basis, periodogrammed, and plotted.