Abstract:Using the N-body simulations of the AEMULUS Project, we construct an emulator for the non-linear clustering of galaxies in real and redshift space. We construct our model of galaxy bias using the halo occupation framework, accounting for possible velocity bias. The model includes 15 parameters, including both cosmological and galaxy bias parameters. We demonstrate that our emulator achieves ∼ 1% precision at the scales of interest, 0.1 < r < 10 h −1 Mpc, and recovers the true cosmology when tested against inde… Show more
“…First, it is important to realise that a surrogate model such as Gaussian process emulation cannot perfectly reproduce the full forward-modelling approach. In the case of Zhai et al (2019), it was shown that emulator inaccuracies are roughly of the same order of magnitude as the typical observational uncertainties of the data. This has two important implications.…”
Section: Surrogate Modelmentioning
confidence: 87%
“…Finally, the predictionsD are used as training points for an emulator, most commonly a Gaussian process emulator, to predictD for arbitrary points in the cosmology and galaxy-halo parameter space without the need to re-run expensive simulations. The most extensive example of such an emulator approach in the context of the small-scale clustering of galaxies is the work of Zhai et al (2019) (also see Kwan et al 2015;Nishimichi et al 2018;Wibking et al 2019b). In this study, the authors constructed an emulator for the redshift-space clustering of galaxies in the BOSS CMASS survey.…”
Section: Surrogate Modelmentioning
confidence: 99%
“…First, these emulator inaccuracies will degrade or possibly bias our posterior inference. In the work of Zhai et al (2019), typical cosmological constraints are degraded by a factor of ∼ 1.5 due to emulator noise. Secondly, the emulator accuracy will likely degrade further with increased dimensionality of C G. In principle, this could be mitigated by increasing the number of training points for the emulator.…”
Section: Surrogate Modelmentioning
confidence: 99%
“…Then, Z(D|C) is a R n C → R and D(C, G) a R n C +n G → R n D function. Typical numbers of (nD, nC, nG) are (42, 5, 12) (More et al 2015) or (37, 8, 7) (Zhai et al 2019). In principle, nD can be reduced in specific cases through principal component analysis (Nishimichi et al 2018) or by emulating fitting functions (McClintock et al 2019).…”
Section: Cosmological Evidencementioning
confidence: 99%
“…In other words, the uncertainty in the mean of the posterior of f σ8 due to the finite sampling of cosmology is of order 4 times smaller than the posterior uncertainty in f σ8 itself. Thus, our approach might be more precise than the emulation method which adds significant noise due to emulator errors (Zhai et al 2019).…”
Extracting accurate cosmological information from galaxy-galaxy and galaxy-matter correlation functions on non-linear scales ( < ∼ 10 h −1 Mpc) requires cosmological simulations. Additionally, one has to marginalise over several nuisance parameters of the galaxy-halo connection. However, the computational cost of such simulations prohibits naive implementations of stochastic posterior sampling methods like Markov chain Monte Carlo (MCMC) that would require of order O(10 6 ) samples in cosmological parameter space. Several groups have proposed surrogate models as a solution: a so-called emulator is trained to reproduce observables for a limited number of realisations in parameter space. Afterwards, this emulator is used as a surrogate model in an MCMC analysis. Here, we demonstrate a different method called Cosmological Evidence Modelling (CEM). First, for each simulation, we calculate the Bayesian evidence marginalised over the galaxy-halo connection by repeatedly populating the simulation with galaxies. We show that this Bayesian evidence is directly related to the posterior probability of cosmological parameters. Finally, we build a physically motivated model for how the evidence depends on cosmological parameters as sampled by the simulations. We demonstrate the feasibility of CEM by using simulations from the Aemulus simulation suite and forecasting cosmological constraints from BOSS CMASS measurements of redshift-space distortions. Our analysis includes an exploration of how galaxy assembly bias affects cosmological inference. Overall, CEM has several potential advantages over the more common approach of emulating summary statistics, including the ability to easily marginalise over highly complex models of the galaxy-halo connection and greater accuracy, thereby reducing the number of simulations required.
“…First, it is important to realise that a surrogate model such as Gaussian process emulation cannot perfectly reproduce the full forward-modelling approach. In the case of Zhai et al (2019), it was shown that emulator inaccuracies are roughly of the same order of magnitude as the typical observational uncertainties of the data. This has two important implications.…”
Section: Surrogate Modelmentioning
confidence: 87%
“…Finally, the predictionsD are used as training points for an emulator, most commonly a Gaussian process emulator, to predictD for arbitrary points in the cosmology and galaxy-halo parameter space without the need to re-run expensive simulations. The most extensive example of such an emulator approach in the context of the small-scale clustering of galaxies is the work of Zhai et al (2019) (also see Kwan et al 2015;Nishimichi et al 2018;Wibking et al 2019b). In this study, the authors constructed an emulator for the redshift-space clustering of galaxies in the BOSS CMASS survey.…”
Section: Surrogate Modelmentioning
confidence: 99%
“…First, these emulator inaccuracies will degrade or possibly bias our posterior inference. In the work of Zhai et al (2019), typical cosmological constraints are degraded by a factor of ∼ 1.5 due to emulator noise. Secondly, the emulator accuracy will likely degrade further with increased dimensionality of C G. In principle, this could be mitigated by increasing the number of training points for the emulator.…”
Section: Surrogate Modelmentioning
confidence: 99%
“…Then, Z(D|C) is a R n C → R and D(C, G) a R n C +n G → R n D function. Typical numbers of (nD, nC, nG) are (42, 5, 12) (More et al 2015) or (37, 8, 7) (Zhai et al 2019). In principle, nD can be reduced in specific cases through principal component analysis (Nishimichi et al 2018) or by emulating fitting functions (McClintock et al 2019).…”
Section: Cosmological Evidencementioning
confidence: 99%
“…In other words, the uncertainty in the mean of the posterior of f σ8 due to the finite sampling of cosmology is of order 4 times smaller than the posterior uncertainty in f σ8 itself. Thus, our approach might be more precise than the emulation method which adds significant noise due to emulator errors (Zhai et al 2019).…”
Extracting accurate cosmological information from galaxy-galaxy and galaxy-matter correlation functions on non-linear scales ( < ∼ 10 h −1 Mpc) requires cosmological simulations. Additionally, one has to marginalise over several nuisance parameters of the galaxy-halo connection. However, the computational cost of such simulations prohibits naive implementations of stochastic posterior sampling methods like Markov chain Monte Carlo (MCMC) that would require of order O(10 6 ) samples in cosmological parameter space. Several groups have proposed surrogate models as a solution: a so-called emulator is trained to reproduce observables for a limited number of realisations in parameter space. Afterwards, this emulator is used as a surrogate model in an MCMC analysis. Here, we demonstrate a different method called Cosmological Evidence Modelling (CEM). First, for each simulation, we calculate the Bayesian evidence marginalised over the galaxy-halo connection by repeatedly populating the simulation with galaxies. We show that this Bayesian evidence is directly related to the posterior probability of cosmological parameters. Finally, we build a physically motivated model for how the evidence depends on cosmological parameters as sampled by the simulations. We demonstrate the feasibility of CEM by using simulations from the Aemulus simulation suite and forecasting cosmological constraints from BOSS CMASS measurements of redshift-space distortions. Our analysis includes an exploration of how galaxy assembly bias affects cosmological inference. Overall, CEM has several potential advantages over the more common approach of emulating summary statistics, including the ability to easily marginalise over highly complex models of the galaxy-halo connection and greater accuracy, thereby reducing the number of simulations required.
We review the field of collisionless numerical simulations for the large-scale structure of the Universe. We start by providing the main set of equations solved by these simulations and their connection with General Relativity. We then recap the relevant numerical approaches: discretization of the phase-space distribution (focusing on N-body but including alternatives, e.g., Lagrangian submanifold and Schrödinger–Poisson) and the respective techniques for their time evolution and force calculation (direct summation, mesh techniques, and hierarchical tree methods). We pay attention to the creation of initial conditions and the connection with Lagrangian Perturbation Theory. We then discuss the possible alternatives in terms of the micro-physical properties of dark matter (e.g., neutralinos, warm dark matter, QCD axions, Bose–Einstein condensates, and primordial black holes), and extensions to account for multiple fluids (baryons and neutrinos), primordial non-Gaussianity and modified gravity. We continue by discussing challenges involved in achieving highly accurate predictions. A key aspect of cosmological simulations is the connection to cosmological observables, we discuss various techniques in this regard: structure finding, galaxy formation and baryonic modelling, the creation of emulators and light-cones, and the role of machine learning. We finalise with a recount of state-of-the-art large-scale simulations and conclude with an outlook for the next decade.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.