This repo contains a complimentary code for the GWAK LIGO paper.
To start, install Miniconda
(if you don't already have it) to be able to create a virtual environment.
Once conda is there, setup a virtual environment using the file from the repo
conda env create --file environment.yaml
conda activate gwak
Install additional package inside the create virtual environment
pip install snakemake
pip install matplotlib-label-lines
To make pytorch available, do
pip install torch==2.0.1
Good thing to do before running (only has to be done once) is to create
an output/ directory that is a symbolic link to your work space (
wherever you have a lot of available space), to be able to store all the data.
ln -s {your-work-path-to-output} output/
Otherwise Snakemake will create output/ directory in the same location this repo is.
Now every time you want to run the code, first activate the virtual environment
conda activate gwak
The code is organized using Snakemake.
The Snakemake workflow management system is a tool to create reproducible and scalable data analyses.
To run snakemake do
snakemake -c1 {rule_name}
where -c1 specifies number of cores provided (one in this case).
It became required to specify it in the latest versions of snakemake,
so to make life easier you can add
alias snakemake="snakemake -c1" to your bash/zsch/whatever profile
and afterwards simply run snakemake {rule_name}.
If you want to run a rule, but Snakemake tells you Nothing to be done, use -f
to force it. Use -F to also force all the upstream rules to be re-run.
In order to be able to submit jobs to HTCondor, install snakemake-condor-profile.
Sending Snakemake process to HTCondor:
$ snakemake --profile HTCondor
Each step of the analysis is represented as a corresponding rule in the Snakefile.
There are several steps starting with downloading and generating data samples,
pre-processing them and training the GWAK space using these samples.
Followed by the Pearson correlation calculation and the evaluation step.
The details of all the available rules are given below.
All the parameters of generation and training are specified in config.py, that's
the only place they are specified and the only place through which they can be changed.
The analysis pipeline starts with fetching valid 03a background data and running Omicron software on it to select loud and quite segments of data taking.
To find the valid data segments, we have in the repo under the data/ folder valid segments for H1 and L1 downloaded from the LIGO website.
snakemake -c1 run_omicron
Next, the LIGO data is fetched and saved with
snakemake -c1 fetch_data
To generate all the necessary samples for the analysis from downloaded data use generate_dataset rule.
These rules are written using Wildcards, such that one rule can be used for
each of four datasets, eg BBH, SG, glitches and background.
Finally, all the plots from the paper can be reproduced with the plotting rule
snakemake -c1 plot_results