cosmicjellyfish

cosmicjellyfish is a Python package for cosmological Fisher forecasts. It helps prepare and compare forecasts for surveys such as Euclid, DESI, SKAO and CMB experiments.

Validation data

Published validation results, posterior chains and HPC workflows live in the public cjfp-data repository on DagsHub. See the download and workflow guide for targeted DVC downloads and instructions to run the comparisons.

Features

• Observables. Photometric (GCph, WL) and spectroscopic (GCsp) galaxy clustering and weak lensing, CMB (CMB_T, CMB_E, CMB_B) and HI intensity mapping (IM).
• Backends. CAMB, CLASS, a built-in symbolic backend that needs no Einstein-Boltzmann solver, optional JAXcosmo, or precomputed tables.
• Derivatives. Four finite-difference schemes (3PT, STEM, POLY, 4PT_FWD), chosen with one option.
• Automatic differentiation. With the optional jaxcosmo extra, photometric and spectroscopic observables are differentiated with JAX, and can be compared against finite differences.
• Analysis and plots. Marginalise, reshuffle and combine Fisher matrices, and draw triangle plots to compare surveys and settings.
• Likelihood sampling. Sample the full posterior of the photometric and spectroscopic likelihoods with Nautilus and compare it with the Fisher forecast.

Installation

cosmicjellyfish supports Python >=3.10,<3.13.

pip install cosmicjellyfish

The symbolic backend additionally needs symbolic_pofk, which is not on PyPI. Install it with

pip install git+https://github.com/DeaglanBartlett/symbolic_pofk.git

To work from a source checkout, use uv:

git clone https://github.com/santiagocasas/cosmicjellyfish.git
cd cosmicjellyfish
uv sync --extra dev

Prefer mamba/conda? From a checkout, mamba env create -f environment.yml builds a conda-forge environment called cosmicjellyfish with everything installed.

Check the installation with

python -c "import cosmicjellyfish; print('cosmicjellyfish import OK')"

and, from a source checkout, run the tests with uv run pytest. Conda setup, optional extras and more are in the installation guide.

Minimal example

A photometric forecast with the symbolic backend (requires symbolic_pofk, see above). It needs no solver setup and runs in a few seconds from any directory, writing its output to results/:

from cosmicjellyfish.fishermatrix import cosmicfish as cff

options = {
    "accuracy": 1,
    "feedback": 1,
    "code": "symbolic",  # or "class" / "camb"
    "cosmo_model": "LCDM",
    "derivatives": "3PT",
    "ell_sampling": 25,  # coarse grid for speed, not production accuracy
    "nonlinear": True,
    "outroot": "first_forecast",
    "results_dir": "results",
    "survey_name": "Euclid",
    "survey_name_photo": "Euclid-Photometric-ISTF-Pessimistic",
    "survey_name_spectro": False,
}

fiducial = {"Omegam": 0.32, "h": 0.67}
freepars = {"Omegam": 0.01, "h": 0.01}  # relative finite-difference steps

forecast = cff.FisherMatrix(
    fiducialpars=fiducial,
    freepars=freepars,
    options=options,
    observables=["WL", "GCph"],
    cosmoModel=options["cosmo_model"],
    surveyName=options["survey_name"],
)

fisher = forecast.compute()
print(fisher.fisher_matrix.shape, fisher.get_param_names())

Documentation

• Documentation: getting started, user guide, tutorials, API reference and development notes.
• Validation dashboard
• Citation metadata (CITATION.cff)

To build the documentation locally:

uv sync --extra docs
make docs   # live-reloading server at http://127.0.0.1:8000

CI builds the docs with warnings treated as errors:

uv run --extra docs sphinx-build -W --keep-going -b html docs/source docs/build/html

Contributors ✨

Thanks goes to these wonderful people (emoji key):

Sefa Pamuk 💻 🎨 🤔 ⚠️ 🔬 👀 📓 💡 📖 🖋 🐛

Santiago Casas 🐛 🖋 🔣 📖 💡 🤔 🚧 📆 🔬 👀 ⚠️ ✅ 📓

Matteo Martinelli 💻 🎨 🤔

Sabarish 🤔 💻 📖

This project follows the all-contributors specification. Contributions of any kind are welcome.