Vision transformer models segment proto-halo regions in initial cosmological density fields by final mass at z=0, outperforming both CNNs and the PINOCCHIO perturbation-theory code.
Multiscale Gaussian Random Fields for Cosmological Simulations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
This paper describes the generation of initial conditions for numerical simulations in cosmology with multiple levels of resolution, or multiscale simulations. We present the theory of adaptive mesh refinement of Gaussian random fields followed by the implementation and testing of a computer code package performing this refinement called GRAFIC2. This package is available to the computational cosmology community at http://arcturus.mit.edu/grafic/ or by email from the author.
fields
astro-ph.CO 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Segmenting proto-halos with vision transformers
Vision transformer models segment proto-halo regions in initial cosmological density fields by final mass at z=0, outperforming both CNNs and the PINOCCHIO perturbation-theory code.