Parameter estimation for the cosmic microwave background with Bayesian neural networks

Hortúa, Héctor J.; Volpi, Riccardo; Marinelli, Dimitri; Malagò, Luigi

doi:10.1103/physrevd.102.103509

Cited by 30 publications

(27 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We certainly need to better understand the predictive uncertainty. This may include by dividing it into epistemic uncertainty and aleatoric uncertainty (Kiureghian & Ditlevsen 2009;Gal 2016;Hortúa et al 2020). The epistemic uncertainty is reducible by increasing the amount of observations as the limited training set can be insufficient for the entire feature space, while aleatoric uncertainty captures the noise of intrinsic randomness (such as photon noise) and cannot be reduced by collecting more data.…”

Section: Discussionmentioning

confidence: 99%

Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks

Hu,

Chen,

Wang

2022

Preprint

View full text Add to dashboard Cite

We present a data-driven method based on Long Short-Term Memory (LSTM) neural networks to analyze spectral time series of Type Ia Supernovae (SNe Ia). The dataset includes 3091 spectra from 361 individual SNe Ia. The method allows for accurate reconstruction of the spectral sequence of an SN Ia based on a single observed spectrum around maximum light. The precision of the spectral reconstruction increases with more spectral time coverages, but significant benefit of multiple epoch data at around optical maximum is only evident for observations separated by more than a week. The method shows great power in extracting the spectral information of SNe Ia, and suggests that the most critical information of an SN Ia can be derived from a single spectrum around the optical maximum. The algorithm we have developed is important for the planning of spectroscopic follow-up observations of future supernova surveys with the LSST/Rubin and the WFIRST/Roman telescopes.

show abstract

Section: Discussionmentioning

confidence: 99%

Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks

Hu,

Chen,

Wang

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Alternatively, ResUNet-CMB could be used as the core component to construct a Bayesian neural network [92] that is capable of producing a probabilistic distribution of outputs. Bayesian neural net-works have a growing presence in cosmology with applications such as parameter and uncertainty estimation with the CMB [93,94] and with strong gravitational lensing [95,96]. Incorporating ResUNet-CMB into a Bayesian neural network would allow for predictions with well-characterized posterior probabilities.…”

Section: Discussionmentioning

confidence: 99%

Reconstructing Cosmic Polarization Rotation with ResUNet-CMB

Guzman,

Meyers

2021

Preprint

View full text Add to dashboard Cite

Cosmic polarization rotation, which may result from parity-violating new physics or the presence of primordial magnetic fields, converts E-mode polarization of the cosmic microwave background (CMB) into B-mode polarization. Anisotropic cosmic polarization rotation leads to statistical anisotropy in CMB polarization and can be reconstructed with quadratic estimator techniques similar to those designed for gravitational lensing of the CMB. At the sensitivity of upcoming CMB surveys, lensing-induced B-mode polarization will act as a limiting factor in the search for anisotropic cosmic polarization rotation, meaning that an analysis which incorporates some form of delensing will be required to improve constraints on the effect with future surveys. In this paper we extend the ResUNet-CMB convolutional neural network to reconstruct anisotropic cosmic polarization rotation in the presence of gravitational lensing and patchy reionization, and we show that the network simultaneously reconstructs all three effects with variance that is lower than that from the standard quadratic estimator nearly matching the performance of an iterative reconstruction method.

show abstract

“…Assuming the variational distribution can be expressed as a mean plus a perturbation, by randomly multiplying each perturbation by either {1, −1} one can ensure that the weights across a batch are at least partially decorrelated. This method has proven to be effective in recent applications of BNNs [42,43], with the additional advantage of being available as pre-built implementations in popular Deep Learning libraries like TensorFlow 1 for both dense and convolutional layers [61]. In this paper we make use of TensorFlow [62] and TensorFlow Probability [61] throughout.…”

Section: Classification In Bnnsmentioning

confidence: 99%

“…Once a probability distribution is defined, BNNs are also able to determine whether an example does not belong to any of the classes in the training set. BNNs have recently been applied in many fields such as gravitational waves [41,42], the Cosmic Microwave Background [43], autonomous driving [44] and cellular image classification [45].…”

Section: Introductionmentioning

confidence: 99%

Seeking New Physics in Cosmology with Bayesian Neural Networks: Dark Energy and Modified Gravity

Mancarella¹,

Kennedy²,

Bose³

et al. 2020

Preprint

View full text Add to dashboard Cite

We study the potential of Bayesian Neural Networks (BNNs) to detect new physics in the dark matter power spectrum, concentrating here on evolving dark energy and modifications to General Relativity. After introducing a new technique to quantify classification uncertainty in BNNs, we train two BNNs on mock matter power spectra produced using the publicly available code ReACT in the k-range (0.01 − 2.5) hMpc −1 and redshift bins (0.1, 0.478, 0.783, 1.5) with Euclid-like noise. The first network classifies spectra into five labels including ΛCDM, f (R), wCDM, Dvali-Gabadaze-Porrati (DGP) gravity and a "random" class whereas the second is trained to distinguish ΛCDM from non-ΛCDM. Both networks achieve a comparable training, validation and test accuracy of ∼ 95%. Each network is also capable of detecting deviations from ΛCDM that were not included in the training set, demonstrated with spectra generated using the growth-index γ. We then quantify the constraining power of each network by computing the smallest deviation from ΛCDM such that the noise-averaged non-ΛCDM classification probability is at least 2σ, finding these bounds to be f R0 10 −7 , Ω rc 10 −2 , −1.05 w 0 0.95, −0.2 w a 0.2, 0.52 γ 0.59. The bounds on f (R) can be improved by training a specialist network to distinguish solely between ΛCDM and f (R) power spectra which can detect a non-zero f R0 at O 10 −8 with a confidence > 2σ. We expect that further developments, such as the inclusion of smaller length scales or additional extensions to ΛCDM, will only improve the potential of BNNs to detect new physics using cosmological datasets. Alongside this paper we publish the publicly available code Bayesian Cosmological Network (BaCoN) which can be accessed at the github repository https://github.com/Mik3M4n/BaCoN with the training and test data available at https://doi.org/10.5281/zenodo.4309918.

show abstract

Parameter estimation for the cosmic microwave background with Bayesian neural networks

Cited by 30 publications

References 54 publications

Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks

Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks

Reconstructing Cosmic Polarization Rotation with ResUNet-CMB

Seeking New Physics in Cosmology with Bayesian Neural Networks: Dark Energy and Modified Gravity

Contact Info

Product

Resources

About