Robust marginalization of baryonic effects for cosmological inference at the field level

Villaescusa-Navarro, Francisco; Genel, Shy; Anglés-Alcázar, Daniel; Spergel, David N.; Li, Yin; Wandelt, B. D.; Thiele, Leander; Nicola, Andrina; Matilla, José Manuel Zorrilla; Shao, Helen; Hassan, Sultan; Narayanan, Desika; Davé, Romeel; Vogelsberger, Mark

doi:10.48550/arxiv.2109.10360

Cited by 30 publications

(46 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One could use (3.15) as the likelihood for the observed map of galaxies (suitably generalized to redshift space [101]) and solve the high-dimensional minimization problem by brute force. Of course, good approximate methods exist for these types of problems (using machine learning [53][54][55][56][57][58][59][60]) and this approach is potentially the best way to perform a map-level analysis in practice. However, as our interest is in understanding the nature of the cosmic information in the maps, we will take the following analytic approach: we will first guess a model of the initial Gaussian map, δguess ( x), and find the maximum likelihood values of the parameters b n and f eq NL for the observed map, δ obs g ( x), while holding δguess ( x) fixed.…”

Section: With Cosmic Variancementioning

confidence: 99%

“…Several groups have been using field-level likelihoods of the dark matter, halos and galaxies in order to test our modeling beyond the predictions of individual correlation functions [47][48][49][50][51]. In parallel, there is a large effort to apply "simulation-based inference" (made feasible by the application of machine learning) to cosmological data analysis [52][53][54][55][56][57][58][59][60]. By forward modeling the cosmological maps, these approaches try to sample realizations of the initial conditions (and the cosmological parameters) and find those that best reproduce the observed maps of the late universe.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The Power of Locality: Primordial Non-Gaussianity at the Map Level

Baumann,

Green

2021

Preprint

View full text Add to dashboard Cite

Primordial non-Gaussianity is a sensitive probe of the inflationary era, with a number of important theoretical targets living an order of magnitude beyond the reach of current CMB constraints. Maps of the large-scale structure of the universe, in principle, have the raw statistical power to reach these targets, but the complications of nonlinear evolution are thought to present serious, if not insurmountable, obstacles to reaching these goals. In this paper, we will argue that the challenge presented by nonlinear structure formation has been overstated. The information encoded in primordial non-Gaussianity resides in nonlocal correlations of the density field at three or more points separated by cosmological distances. In contrast, nonlinear evolution only alters the density field locally and cannot create or destroy these long-range correlations. This locality property of the late-time non-Gaussianity is obscured in Fourier space and in the standard bispectrum searches for primordial non-Gaussianity. We therefore propose to measure non-Gaussianity in the position space maps of the large-scale structure. As a proof of concept, we study the case of equilateral non-Gaussianity, for which the degeneracy with late-time nonlinearities is the most severe. We show that a map-level analysis is capable of breaking this degeneracy and thereby significantly improve the constraining power over previous estimates. Our findings suggest that "simulation-based inference" involving the forward modeling of large-scale structure maps has the potential to dramatically impact the search for primordial non-Gaussianity.

show abstract

Section: With Cosmic Variancementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

The Power of Locality: Primordial Non-Gaussianity at the Map Level

Baumann,

Green

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Another avenue is to use machine learning techniques, e.g. neural networks, to find an approximation to the optimal estimator (Ravanbakhsh et al 2017;Schmelzle et al 2017;Gupta et al 2018;Ribli et al 2019;Fluri et al 2019;Ntampaka et al 2019;Hassan et al 2020;Zorrilla Matilla et al 2020;Villaescusa-Navarro et al 2021a;Lu et al 2021). Recent works have shown that even for fields that are very contaminated by astrophysical effects, it is possible to extract cosmological information from small scales (Villaescusa-Navarro et al 2021b).…”

Section: Introductionmentioning

confidence: 99%

Cosmology with one galaxy?

Villaescusa-Navarro,

Ding,

Genel

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Galaxies can be characterized by many internal properties such as stellar mass, gas metallicity, and star-formation rate. We quantify the amount of cosmological and astrophysical information that the internal properties of individual galaxies and their host dark matter halos contain. We train neural networks using hundreds of thousands of galaxies from 2,000 state-of-the-art hydrodynamic simulations with different cosmologies and astrophysical models of the CAMELS project to perform likelihood-free inference on the value of the cosmological and astrophysical parameters. We find that knowing the internal properties of a single galaxy allow our models to infer the value of Ω m , at fixed Ω b , with a ∼ 10% precision, while no constraint can be placed on σ 8 . Our results hold for any type of galaxy, central or satellite, massive or dwarf, at all considered redshifts, z ≤ 3, and they incorporate uncertainties in astrophysics as modeled in CAMELS. However, our models are not robust to changes in subgrid physics due to the large intrinsic differences the two considered models imprint on galaxy properties. We find that the stellar mass, stellar metallicity, and maximum circular velocity are among the most important galaxy properties to determine the value of Ω m . We believe that our results can be explained taking into account that changes in the value of Ω m , or potentially Ω b /Ω m , affect the dark matter content of galaxies. That effect leaves a distinct signature in galaxy properties to the one induced by galactic processes. Our results suggest that the low-dimensional manifold hosting galaxy properties provides a tight direct link between cosmology and astrophysics.

show abstract

“…A useful meta-sample comes from the CAMELS project (Villaescusa- Navarro et al 2021cNavarro et al , 2022a, a suite of 2, 000 magneto-hydrodynamic simulations with varying initial conditions, cosmological and baryonic parameters, and two different subgrid physics models, designed with machine learning applications in mind (e.g. Villaescusa- Navarro et al 2021b). The CAMELS boxes are however small (25 Mpc/ℎ in side length), increasing sample variance.…”

Section: Simulations and Featuresmentioning

confidence: 99%

The scatter in the galaxy-halo connection: a machine learning analysis

Stiskalek,

Bartlett,

Desmond

et al. 2022

Preprint

View full text Add to dashboard Cite

We apply machine learning, a powerful method for uncovering complex correlations in high-dimensional data, to the galaxyhalo connection of cosmological hydrodynamical simulations. The mapping between galaxy and halo variables is stochastic in the absence of perfect information, but conventional machine learning models are deterministic and hence cannot capture its intrinsic scatter. To overcome this limitation, we design an ensemble of neural networks with a Gaussian loss function that predict probability distributions, allowing us to model statistical uncertainties in the galaxy-halo connection as well as its best-fit trends. We extract a number of galaxy and halo variables from the Horizon-AGN and IllustrisTNG100-1 simulations and quantify the extent to which knowledge of some subset of one enables prediction of the other. This allows us to identify the key features of the galaxy-halo connection and investigate the origin of its scatter in various projections. We find that while halo properties beyond mass account for up to 50 per cent of the scatter in the halo-to-stellar mass relation, the prediction of stellar half-mass radius or total gas mass is not substantially improved by adding further halo properties. We also use these results to investigate semi-analytic models for galaxy size in the two simulations, finding that assumptions relating galaxy size to halo size or spin are not successful.

show abstract

Robust marginalization of baryonic effects for cosmological inference at the field level

Cited by 30 publications

References 15 publications

The Power of Locality: Primordial Non-Gaussianity at the Map Level

The Power of Locality: Primordial Non-Gaussianity at the Map Level

Cosmology with one galaxy?

The scatter in the galaxy-halo connection: a machine learning analysis

Contact Info

Product

Resources

About