CosmoFlow: Scale-Aware Representation Learning for Cosmology with Flow Matching

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kannan, Sidharth, Qiu, Tian, Cuesta-Lazaro, Carolina, Jeong, Haewon
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912484837818368
author Kannan, Sidharth
Qiu, Tian
Cuesta-Lazaro, Carolina
Jeong, Haewon
author_facet Kannan, Sidharth
Qiu, Tian
Cuesta-Lazaro, Carolina
Jeong, Haewon
contents Generative machine learning models have been demonstrated to be able to learn low dimensional representations of data that preserve information required for downstream tasks. In this work, we demonstrate that flow matching based generative models can learn compact, semantically rich latent representations of field level cold dark matter (CDM) simulation data without supervision. Our model, CosmoFlow, learns representations 32x smaller than the raw field data, usable for field level reconstruction, synthetic data generation, and parameter inference. Our model also learns interpretable representations, in which different latent channels correspond to features at different cosmological scales.
format Preprint
id arxiv_https___arxiv_org_abs_2507_11842
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CosmoFlow: Scale-Aware Representation Learning for Cosmology with Flow Matching
Kannan, Sidharth
Qiu, Tian
Cuesta-Lazaro, Carolina
Jeong, Haewon
Cosmology and Nongalactic Astrophysics
Machine Learning
Generative machine learning models have been demonstrated to be able to learn low dimensional representations of data that preserve information required for downstream tasks. In this work, we demonstrate that flow matching based generative models can learn compact, semantically rich latent representations of field level cold dark matter (CDM) simulation data without supervision. Our model, CosmoFlow, learns representations 32x smaller than the raw field data, usable for field level reconstruction, synthetic data generation, and parameter inference. Our model also learns interpretable representations, in which different latent channels correspond to features at different cosmological scales.
title CosmoFlow: Scale-Aware Representation Learning for Cosmology with Flow Matching
topic Cosmology and Nongalactic Astrophysics
Machine Learning
url https://arxiv.org/abs/2507.11842