veloce

Welcome to veloce, the velocity power spectrum covariance emulator! We use an emulator based on neural networks to accelerate the prediction of covariance matrices for different cosmologies in the context of supernovae studies. If you are interested and want to know more, check out the paper, raise an issue or contact Davide Piras.

Installation

To use the emulator and/or sample your supernovae posterior, follow these steps:

1. (optional) conda create -n veloce python=3.11 jupyter ipython=8.20 (create a custom conda environment with python 3.11)

2. (optional) conda activate veloce (activate it)

3. Install the package:

    pip install velocemu
    python -c 'import velocemu'
   

   or alternatively, clone the repository and install it:

    git clone https://github.com/dpiras/veloce.git
    cd veloce
    pip install . 
    python -c 'import velocemu'
   

Usage

Cloning the repository will also give you access to all Jupyter notebooks, which include information on how to generate a single element of the covariance, use the emulator, and sample the posterior.

Trained models

You can find the available models in this folder, which will be updated when new models become available. Currently, we provide the model that leads to the final results of the paper, namely the nonlinear case with fixed $\sigma_{\rm u}$, but more models are in production. If you are interested in other models, please reach out or contact Davide Piras. Also note that it should be straightforward for you to train your own models using CosmoPower, and then add them under velocemu/trained_models.

Contributing and contacts

Feel free to fork this repository to work on it; otherwise, please raise an issue or contact Davide Piras.

Citation

If you use veloce, please cite the corresponding paper:

 @ARTICLE{Piras25,
      author = {{Piras}, Davide and {Sorrenti}, Francesco and {Durrer}, Ruth and {Kunz}, Martin},
      title = "{Anchors no more: Using peculiar velocities to constrain $H_0$ and the primordial Universe without calibrators}",
      journal = {arXiv e-prints},
      year = 2025,
      month = apr,
      eid = {arXiv:2504.10453},
      pages = {arXiv:2504.10453},
      doi = {10.48550/arXiv.2504.10453},
      archivePrefix = {arXiv},
      eprint = {2504.10453},
      primaryClass = {astro-ph.CO},
      adsurl = {https://ui.adsabs.harvard.edu/abs/2025arXiv250410453P},
 }

License

veloce is released under the GPL-3 license (see LICENSE) subject to the non-commercial use condition.

 veloce
 Copyright (C) 2025 Davide Piras & contributors

 This program is released under the GPL-3 license (see LICENSE.txt), subject to a non-commercial use condition.