mardom/variational-diffusion-cdm

Diffusion generative model for DM reconstruction

Diffusion generative model for reconstruction of dark matter fields from galaxies

Victoria Ono, Core Francisco Park, Nayantara Mudur et al.

Abstract

We develop a diffusion generative model to reconstruct dark matter fields from galaxies. The diffusion model is trained on the CAMELS simulation suite that contains thousands of state-of-the-art galaxy formation simulations with varying cosmological parameters and sub-grid astrophysics. We demonstrate that the diffusion model can predict the unbiased posterior distribution of the underlying dark matter fields from the given stellar mass fields, while being able to marginalize over uncertainties in cosmological and astrophysical models.

Dataset

The CAMELS Multifield Dataset used to train the model can be found here.

Code overview

• The diffusion model is defined in model/vdm_model.py, with auxiliary utilities (noise schedules, drawing figures in a logger) in model/utils/utils.py. The model is based on the google-research/vdm repo.
• The score model is called from model/networks.py.
• The data directory contains files for constructing the PyTorch DataLoaders for the training datasets.
• The notebooks directory contains notebooks used to produce results for the paper.

Citation

If you use this code, please cite our paper:

@misc{ono2024debiasing,
      title={Debiasing with Diffusion: Probabilistic reconstruction of Dark Matter fields from galaxies with CAMELS}, 
      author={Victoria Ono and Core Francisco Park and Nayantara Mudur and Yueying Ni and Carolina Cuesta-Lazaro and Francisco Villaescusa-Navarro},
      year={2024},
      eprint={2403.10648},
      archivePrefix={arXiv},
      primaryClass={astro-ph.CO}
}