CosmoFlow: Scale-Aware Representation Learning for Cosmology with Flow Matching
This work provides a domain-specific tool for cosmology by enabling efficient analysis and generation of simulation data, though it is incremental as it applies flow matching to a known bottleneck in the field.
The authors tackled the problem of learning compact, semantically rich latent representations from cold dark matter simulation data without supervision, achieving a 32x reduction in size while enabling tasks like field reconstruction and parameter inference.
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.