The CAMELS Multifield Data Set: Learning the Universe’s Fundamental Parameters with Artificial Intelligence

Villaescusa-Navarro, Francisco; Genel, Shy; Anglés-Alcázar, Daniel; Thiele, Leander; Davé, Romeel; Narayanan, Desika; Nicola, Andrina; Li, Yin; Villanueva-Domingo, Pablo; Wandelt, B. D.; Spergel, David N.; Somerville, Rachel S.; Matilla, José Manuel Zorrilla; Mohammad, Faizan G.; Hassan, Sultan; Shao, Helen; Wadekar, Digvijay; Eickenberg, Michael; Wong, Kaze W. K.; Contardo, Gabriella; Jo, Yongseok; Moser, Emily; Lau, Erwin T.; Valle, Luis Fernando Machado Poletti; Perez, Lucia A.; Nagai, Daisuke; Battaglia, Nicholas; Vogelsberger, Mark

doi:10.3847/1538-4365/ac5ab0

Cited by 37 publications

(27 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project [17][18][19] consists of 4,233 numerical simulations which comprise dark matter only and stateof-the-art hydrodynamic simulations. The latter consider two different implementations of the subgrid physics models, namely, IllustrisTNG [20] and SIMBA [21].…”

Section: Data and Modelmentioning

confidence: 99%

“…(Mgas), neutral hydrogen density (HI), magnetic field (B) and gas metallicity (Z). The fields have been selected based on how sensitive they are to the variation of the parameters of interest in this study (see [16]). Figure 1 shows the four fields considered in this study in three different realizations which are related to three different sets of cosmological and astrophysical parameters, with different varying feedback strengths.…”

Section: Data and Modelmentioning

confidence: 99%

“…As this set of parameters are highly degenerate (e.g. low A SN2 and high A AGN2 would boost star formation in the outskirts of galaxies and push gas to low dense regions, see Figure 2 in [16]), it is challenging to determine the leading effect. However, the following combination of Ω m = 0.296, σ 8 = 0.963, A SN1 = 0.261, A AGN1 = 0.865, A SN2 = 0.519, A AGN2 = 0.801 (left column) shows the scenario where gas is being pushed out to less dense regions.…”

Section: Data and Modelmentioning

confidence: 99%

“…This construction is such that the aleatoric uncertainty, which is the other component of the total uncertainty, can be estimated. Following the prescription in [16], the optimization (using Adam optimizer) via mini-batch gradient descent is achieved by using the loss function…”

Section: Data and Modelmentioning

confidence: 99%

“…[14] and [15] used similar network architectures to identify reionization sources, and to recontruct the reionization history respectively. For the first time, [16] extracted cosmological and astrophysical parameters from multifield 2D maps, consisting of a set of fields where each one corresponds to a channel at the input of a deep regressor. They showed that their architecture is capable of marginalizing over the contamination by the astrophysical effects at the field level to constrain both the cosmology and astrophysics with a relatively high accuracy.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Predictive uncertainty on improved astrophysics recovery from multifield cosmology

Andrianomena¹,

Hassan²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

We investigate how the constraints on cosmological and astrophysical parameters (Ω m , σ 8 , A SN1 , A SN2 ) vary when exploiting information from multiple fields in cosmology. We make use of a convolutional neural network to retrieve the salient features from different combinations of field maps from IllustrisTNG in the CAMELS project. The fields considered are neutral hydrogen (HI), gas density (Mgas), magnetic fields (B) and gas metallicity (Z). We estimate the predictive uncertainty on the predictions of our model by using Monte Carlo dropout, a Bayesian approximation. Results show that overall, the performance of the model improves on all parameters as the number of channels of its input is increased. As compared to previous works, our model is able to predict the astrophysical parameters with up to 5% higher in accuracy. In the best setup which includes all fields (four channel input, Mgas-HI-B-Z) the model achieves R 2 > 0.96 on all parameters. Similarly, we find that the total uncertainty, which is dominated by the aleatoric uncertainty, decreases as more fields are used to train the model in general. The uncertainties obtained by dropout variational inference are overestimated on all parameters in our case, in that the predictive uncertainty is much larger than the actual squared error. After calibration, which consists of a simple σ scaling method, the average deviation of the total uncertainty from the actual error goes down to 25% at most (on A SN1 ).

show abstract

Section: Data and Modelmentioning

confidence: 99%

Section: Data and Modelmentioning

confidence: 99%

Section: Data and Modelmentioning

confidence: 99%

Section: Data and Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Predictive uncertainty on improved astrophysics recovery from multifield cosmology

Andrianomena¹,

Hassan²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Probabilistic mass-mapping with neural score estimation

Remy

Lanusse

Jeffrey

et al. 2023

A&A

View full text Add to dashboard Cite

Context. Weak lensing mass-mapping is a useful tool for accessing the full distribution of dark matter on the sky, but because of intrinsic galaxy ellipticies, finite fields, and missing data, the recovery of dark matter maps constitutes a challenging, ill-posed inverse problem Aims. We introduce a novel methodology that enables the efficient sampling of the high-dimensional Bayesian posterior of the weak lensing mass-mapping problem, relying on simulations to define a fully non-Gaussian prior. We aim to demonstrate the accuracy of the method to simulated fields, and then proceed to apply it to the mass reconstruction of the HST/ACS COSMOS field. Methods. The proposed methodology combines elements of Bayesian statistics, analytic theory, and a recent class of deep generative models based on neural score matching. This approach allows us to make full use of analytic cosmological theory to constrain the 2pt statistics of the solution, to understand any differences between this analytic prior and full simulations from cosmological simulations, and to obtain samples from the full Bayesian posterior of the problem for robust uncertainty quantification. Results. We demonstrate the method in the κTNG simulations and find that the posterior mean significantly outperfoms previous methods (Kaiser-Squires, Wiener filter, Sparsity priors) both for the root-mean-square error and in terms of the Pearson correlation. We further illustrate the interpretability of the recovered posterior by establishing a close correlation between posterior convergence values and the S/N of the clusters artificially introduced into a field. Finally, we apply the method to the reconstruction of the HST/ACS COSMOS field, which yields the highest-quality convergence map of this field to date. Conclusions. We find the proposed approach to be superior to previous algorithms, scalable, providing uncertainties, and using a fully non-Gaussian prior.

show abstract

Inference of the optical depth to reionization τ from Planck CMB maps with convolutional neural networks

Wolz¹,

Krachmalnicoff²,

Pagano³

2023

A&A

View full text Add to dashboard Cite

The optical depth to reionization, τ, is the least constrained parameter of the cosmological Λ cold dark matter (ΛCDM) model. To date, its most precise value is inferred from large-scale polarized cosmic microwave background (CMB) power spectra from the High Frequency Instrument (HFI) aboard the Planck satellite. These maps are known to contain significant contamination by residual non-Gaussian systematic effects, which are hard to model analytically. Therefore, robust constraints on τ are currently obtained through an empirical cross-spectrum likelihood built from simulations. In this paper, we present a likelihood-free inference of τ from polarized Planck HFI maps which, for the first time, is fully based on neural networks (NNs). NNs have the advantage of not requiring an analytical description of the data and can be trained on state-of-the-art simulations, combining the information from multiple channels. By using Gaussian sky simulations and Planck SRoll2 simulations, including CMB, noise, and residual instrumental systematic effects, we trained, tested, and validated NN models considering different setups. We inferred the value of τ directly from Stokes Q and U maps at ∼4° pixel resolution, without computing angular power spectra. On Planck data, we obtained τNN = 0.0579 ± 0.0082, which is compatible with current EE cross-spectrum results but with a ∼30% larger uncertainty, which can be assigned to the inherent nonoptimality of our estimator and to the retraining procedure applied to avoid biases. While this paper does not improve on current cosmological constraints on τ, our analysis represents a first robust application of NN-based inference on real data, and highlights its potential as a promising tool for complementary analysis of near-future CMB experiments, also in view of the ongoing challenge to achieve the first detection of primordial gravitational waves.

show abstract

The CAMELS Multifield Data Set: Learning the Universe’s Fundamental Parameters with Artificial Intelligence

Cited by 37 publications

References 36 publications

Predictive uncertainty on improved astrophysics recovery from multifield cosmology

Predictive uncertainty on improved astrophysics recovery from multifield cosmology

Probabilistic mass-mapping with neural score estimation

Inference of the optical depth to reionization τ from Planck CMB maps with convolutional neural networks

Contact Info

Product

Resources

About