chenheinrich/Cosmo_Inference

SPHEREx Cosmo_Inference

The SPHEREx Cosmo_Inference is the Inference section of the SPHEREx L4 Cosmology pipeline. It is used for performing simulated likelihood anaysis.

It comprises of the following python packages:

• lss_theory
• spherex_cobaya

We make use of the public code Cobaya as a MCMC sampler.

Before you start

1. Clone the repository:

   git clone https://github.com/chenheinrich/SphereLikes.git cd Cosmo_Inference

2. It is recommended you create a virtual environment before installing the dependencies (Note that Python 2 is decaprecated, we are using Python 3; if you are on a cluster, do module avail python to see what's available and use module load python<X> to select the version that corresponds to Python 3.7):

   python3 -m pip install virtualenv [--user]

   python3 -m venv env

   source env/bin/activate

or conda create -n spherex python=3.7

`conda activate spherex`

We recommend using virtualenv instead of conda, since installing with pip inside a conda environment could cause problems sometimes.

Install Requirements

2. From inside Cosmo_Inference, install packages (spherex_cobaya and lss_theory) in this repository (add --user if you're on a cluster; and remove -e if you are not activately developing these packages):

   python -m pip install -e ./src/lss_theory [--user]

   python -m pip install -e ./src/spherex_cobaya [--user]

3. Test that the packages are properly installed:

   python -c "import lss_theory; import spherex_cobaya"

Run tests

[PLEASE SKIP THIS SECTION, IT IS UNDER CONSTRUCTION]

Current working tests are:

For lss_theory:

1. python -m pytest src/lss_theory/tests

For spherex_cobaya

1. python tests/dev_test_theory.py
2. python -m pytest tests/test_theory.py -m short

Note: Use python -m pytest instead of pytest to ensure that you are using the same pytest you installed earlier with requirements.txt if there are various python versions.

Run scripts from individual packages:

Current directory must be Cosmo_Inference.

Run lss_theory sample scripts:

• Power spectrum signal with RSD (linear theory + FoG)

  python -m lss_theory.scripts.get_ps ./src/lss_theory/sample_inputs/get_ps.yaml

• Bispectrum signal with RSD (linear theory + FoG)

  python -m lss_theory.scripts.get_b3d_rsd ./src/lss_theory/sample_inputs/get_b3d_rsd.yaml

• Bispectrum signal without RSD (linear theory)

  python3 -m lss_theory.scripts.get_b3d_base ./src/lss_theory/sample_inputs/get_b3d_base.yaml

• Bispectrum multipole signal (linear theory)

  python3 -m lss_theory.scripts.get_bis_mult ./src/lss_theory/sample_inputs/get_bis_mult.yaml

• Covariance for Bispectrum with RSD [Not validated; under construction]

  python3 -m lss_theory.scripts.get_covariance_b3d_rsd ./src/lss_theory/sample_inputs/get_covariance_b3d_rsd.yaml

• Covariance for Bispectrum multipole [Not validated; under construction]

  python3 -m lss_theory.scripts.get_covariance_bis_mult ./src/lss_theory/sample_inputs/get_covariance_bis_mult.yaml

• Fisher matrix for Bispectrum with RSD [Not validated; to be relocated]

  python3 -m lss_theory.fisher.fisher_b3d_rsd ./src/lss_theory/sample_inputs/get_fisher_b3d_rsd.yaml

• Fisher matrix for Bispectrum multipoles [Not validated; to be relocated]

  python3 -m lss_theory.fisher.fisher_bis_mult ./src/lss_theory/sample_inputs/get_fisher_bis_mult.yaml

Run spherex_cobaya sample scripts:

`cobaya-run ./src/spherex_cobaya/sample_inputs/cobaya_pars/b3d_rsd_vary_all_but_w0_wa_mnu.yaml -d -f`

To get the data needed for running MCMC chains (covariance matrix and simulated data), use:

[Coming soon]

[Other yaml files are still being upgraded.]

Run sample pipeline:

Current directory must be Cosmo_Inference.

There is a draft pipeline under construction:

./pipeline/scripts/execute.sh

Alternatives: Docker (under development)

[PLEASE SKIP THIS SECTION, THE FOLLOWING NEEDS TO BE UPDATED]

Basic run

After installing Docker, you can pull the image

docker pull chenheinrich/spherex:0.0.1

and run the default chains for SPHEREx (simulated): docker run --rm chenheinrich/spherex:0.0.1

Custom chains

The following is subject to change.

To run a different chain than the basic one, use docker run --rm chenheinrich/spherex:0.0.1 python3 scripts/run_chains.py ./inputs/chains_pars/<run_name>.yaml 1 -d -f -run_in_python replacing <run_name> by the name of your new yaml file. Note that you will need to prepare the elements needed by the chains yourself in this case, by running docker run --rm chenheinrich/spherex:0.0.1 python3 scripts/prep_chains.py ./inputs/chains_pars/<run_name>.yaml.

This will:

1. run the reference cosmology and products needed for computing Alcock-Pazcynski effect (H(z) and D_A(z)).
2. create a covariance matrix with shot noise specified by survey parameters and invert it (which could take a few minutes).
3. create simulated data vector according to the cosmology specified.