arXiv subjects

Natalia Porqueres

Publications and source records attributed to Natalia Porqueres.

15 recordsLinked to original sources

arXiv2026

Nonparametric Variational Inference Reconstruction of the Cosmic Expansion History from SNe Ia -- the charm2 code

Cosmological analyses using the latest set of type Ia SNe data weakly favor an evolving dark energy (EDE) model without strongly disfavoring the standard LCDM paradigm. Nonparametric reconstructions of the expansion history may reveal signal features potentially missed by a parametric LCDM model without laying out a specific functional form for the evolution of dark energy. Information field theory (IFT) is a Bayesian framework for optimal, nonparametric reconstruction algorithms. In this work, we present charm2, the successor to charm1, a previous IFT-based code to reconstruct the cosmic energy density's redshift evolution from SNe Ia. We apply our reconstruction algorithm to the Union2.1, Pantheon+, DESY5 and DESY5-Dovekie data sets to investigate the agreement between the nonparametric reconstruction and the signal suggested by a parametric, flat LCDM model. To enable an accurate Gaussian approximation, we employ geometric variational inference, which finds a coordinate transformation through which a curved posterior gets "flattened". The redshift evolution of the energy density can then be traced on a double-logarithmic scale, which, after de-trending, is well described by a stationary Gaussian process. The nonparametric charm2 reconstructions using the Union2.1 and Pantheon+ data sets are consistent with flat LCDM signal fields. The DESY5 and DESY5-Dovekie reconstructions deviate from flat LCDM comparison fields and are compatible with an evolving dark energy signal. However, using the evidence lower bound (ELBO) measure for model selection, we find no conclusive evidence supporting a preference for non-flat-LCDM features in any of the data sets. We note that at current DESY5 noise levels, the ELBO tends to favor flat LCDM over our nonparametric model although the latter better recovers the ground truth in synthetic EDE data; a trend reversing only at ~7x lower noise covariance.

astro-ph.CO↗

arXiv2026

Differentiable Forward Modeling for Efficient and Accurate Shear Inference

Forthcoming Stage-IV dark energy optical surveys, such as LSST, have the ambitious goal of measuring cosmological parameters at sub-percent precision. Realizing their full scientific potential requires very precise measurement of the cosmic shear signal and control of corresponding systematics. In this work, we present a modern implementation of the Bayesian shear inference framework in Schneider et al. (2015), in the case that the PSF and sky background are known. This framework automatically propagates the pixel-noise measurement error from each galaxy into the final shear estimate, and thus requires no external calibration to handle noise bias. As a first application of this new implementation, we infer the cosmic shear posterior from simulated images consisting of isolated exponential galaxies with semi-realistic noise levels. In this simplified scenario, we estimate the absolute multiplicative bias $|m|$ of our approach to be below $0.9 \times 10^{-3}\,[3\sigma]$ when the intrinsic distribution of galaxy properties is known, and below $1.3 \times 10^{-3}\,[3\sigma]$ when these distributions are inferred alongside shear. Additionally, we make progress towards the algorithm's computational feasibility in the context of modern wide-field surveys, where billions of galaxies must be processed, by leveraging differentiable forward models of galaxies, gradient-based samplers, and GPUs. Our final galaxy-fitting MCMC produces $300$ effective samples of galaxy properties in $0.45$ seconds per galaxy using a single A100 GPU. In the future, we seek to generalize our algorithm to handle selection, detection, and model shear biases so it can be applied to real survey data.

astro-ph.IM↗

arXiv2025

Hybrid Summary Statistics

We present a way to capture high-information posteriors from training sets that are sparsely sampled over the parameter space for robust simulation-based inference. In physical inference problems, we can often apply domain knowledge to define traditional summary statistics to capture some of the information in a dataset. We show that augmenting these statistics with neural network outputs to maximise the mutual information improves information extraction compared to neural summaries alone or their concatenation to existing summaries and makes inference robust in settings with low training data. We introduce 1) two loss formalisms to achieve this and 2) apply the technique to two different cosmological datasets to extract non-Gaussian parameter information.

stat.ML↗

arXiv2024

Hybrid summary statistics: neural weak lensing inference beyond the power spectrum

In inference problems, we often have domain knowledge which allows us to define summary statistics that capture most of the information content in a dataset. In this paper, we present a hybrid approach, where such physics-based summaries are augmented by a set of compressed neural summary statistics that are optimised to extract the extra information that is not captured by the predefined summaries. The resulting statistics are very powerful inputs to simulation-based or implicit inference of model parameters. We apply this generalisation of Information Maximising Neural Networks (IMNNs) to parameter constraints from tomographic weak gravitational lensing convergence maps to find summary statistics that are explicitly optimised to complement angular power spectrum estimates. We study several dark matter simulation resolutions in low- and high-noise regimes. We show that i) the information-update formalism extracts at least $3\times$ and up to $8\times$ as much information as the angular power spectrum in all noise regimes, ii) the network summaries are highly complementary to existing 2-point summaries, and iii) our formalism allows for networks with smaller, physically-informed architectures to match much larger regression networks with far fewer simulations needed to obtain asymptotically optimal inference.

astro-ph.CO↗

arXiv2024

DISCO-DJ I: a differentiable Einstein-Boltzmann solver for cosmology

We present the Einstein-Boltzmann module of the DISCO-DJ (DIfferentiable Simulations for COsmology - Done with JAX) software package. This module implements a fully differentiable solver for the linearised cosmological Einstein-Boltzmann equations in the JAX framework, and allows computing Jacobian matrices of all solver output with respect to all input parameters using automatic differentiation. This implies that along with the solution for a given set of parameters, the tangent hyperplane in parameter space is known as well, which is a key ingredient for cosmological inference and forecasting problems as well as for many other applications. We discuss our implementation and demonstrate that our solver agrees at the per-mille level with the existing non-differentiable solvers CAMB and CLASS, including massive neutrinos and a dark energy fluid with parameterised equation of state. We illustrate the dependence of various summary statistics in large-scale structure cosmology on model parameters using the differentiable solver, and finally demonstrate how it can be easily used for Fisher forecasting, with a forecast for Euclid as an example. Since the implementation is significantly shorter and more modular than existing solvers, we believe it will be more straightforward to extend our solver to include additional physics, such as additional dark energy and dark matter models, modified gravity, or other non-standard physics in the future.

astro-ph.CO↗

arXiv2024

Accuracy requirements on intrinsic alignments for Stage-IV cosmic shear

In the context of cosmological weak lensing studies, intrinsic alignments (IAs) are one the most complicated astrophysical systematic to model, given the poor understanding of the physical processes that cause them. A number of modelling frameworks for IAs have been proposed in the literature, both purely phenomenological or grounded on a perturbative treatment of symmetry-based arguments. However, the accuracy with which any of these approaches is able to describe the impact of IAs on cosmic shear data, particularly on the comparatively small scales ($k\simeq 1\,{\rm Mpc}^{-1}$) to which this observable is sensitive, is not clear. Here we quantify the level of disagreement between the true underlying intrinsic alignments and the theoretical model used to describe them that can be allowed in the context of cosmic shear analyses with future Stage-IV surveys. We consider various models describing this "IA residual", covering both physics-based approaches, as well as completely agnostic prescriptions. The same qualitative results are recovered in all cases explored: for a Stage-IV cosmic shear survey, a mis-modelling of the IA contribution at the $\sim10\%$ level produces shifts of $\lesssim0.5σ$ on the final cosmological parameter constraints. Current and future IA models should therefore aim to achieve this level of accuracy, a prospect that is not unfeasible for models with sufficient flexibility.

astro-ph.CO↗

arXiv2023

Field-level inference of cosmic shear with intrinsic alignments and baryons

We construct a field-based Bayesian Hierarchical Model for cosmic shear that includes, for the first time, the important astrophysical systematics of intrinsic alignments and baryon feedback, in addition to a gravity model. We add to the BORG-WL framework the tidal alignment and tidal torquing model (TATT) for intrinsic alignments and compare them with the non-linear alignment (NLA) model. With synthetic data, we have shown that adding intrinsic alignments and sampling the TATT parameters does not reduce the constraining power of the method and the field-based approach lifts the weak lensing degeneracy. We add baryon effects at the field level using the enthalpy gradient descent (EGD) model. This model displaces the dark matter particles without knowing whether they belong to a halo and allows for self-calibration of the model parameters, which are inferred from the data. We have also illustrated the effects of model misspecification for the baryons. The resulting model now contains the most important physical effects and is suitable for application to data.

astro-ph.CO↗

arXiv2023

LyAl-Net: A high-efficiency Lyman-$α$ forest simulation with a neural network

The inference of cosmological quantities requires accurate and large hydrodynamical cosmological simulations. Unfortunately, their computational time can take millions of CPU hours for a modest coverage in cosmological scales ($\approx (100 {h^{-1}}\,\text{Mpc})^3)$). The possibility to generate large quantities of mock Lyman-$α$ observations opens up the possibility of much better control on covariance matrices estimate for cosmological parameters inference, and on the impact of systematics due to baryonic effects. We present a machine learning approach to emulate the hydrodynamical simulation of intergalactic medium physics for the Lyman-$α$ forest called LyAl-Net. The main goal of this work is to provide highly efficient and cheap simulations retaining interpretation abilities about the gas field level, and as a tool for other cosmological exploration. We use a neural network based on the U-net architecture, a variant of convolutional neural networks, to predict the neutral hydrogen physical properties, density, and temperature. We train the LyAl-Net model with the Horizon-noAGN simulation, though using only 9% of the volume. We also explore the resilience of the model through tests of a transfer learning framework using cosmological simulations containing different baryonic feedback. We test our results by analysing one and two-point statistics of emulated fields in different scenarios, as well as their stochastic properties. The ensemble average of the emulated Lyman-$α$ forest absorption as a function of redshift lies within 2.5% of one derived from the full hydrodynamical simulation. The computation of individual fields from the dark matter density agrees well with regular physical regimes of cosmological fields. The results tested on IllustrisTNG100 showed a drastic improvement in the Lyman-$α$ forest flux without arbitrary rescaling.

astro-ph.CO↗

arXiv2022

The Cosmic Graph: Optimal Information Extraction from Large-Scale Structure using Catalogues

We present an implicit likelihood approach to quantifying cosmological information over discrete catalogue data, assembled as graphs. To do so, we explore cosmological parameter constraints using mock dark matter halo catalogues. We employ Information Maximising Neural Networks (IMNNs) to quantify Fisher information extraction as a function of graph representation. We a) demonstrate the high sensitivity of modular graph structure to the underlying cosmology in the noise-free limit, b) show that graph neural network summaries automatically combine mass and clustering information through comparisons to traditional statistics, c) demonstrate that networks can still extract information when catalogues are subject to noisy survey cuts, and d) illustrate how nonlinear IMNN summaries can be used as asymptotically optimal compressed statistics for Bayesian simulation-based inference. We reduce the area of joint $Ω_m, σ_8$ parameter constraints with small ($\sim$100 object) halo catalogues by a factor of 42 over the two-point correlation function, and demonstrate that the networks automatically combine mass and clustering information. This work utilises a new IMNN implementation over graph data in Jax, which can take advantage of either numerical or auto-differentiability. We also show that graph IMNNs successfully compress simulations away from the fiducial model at which the network is fitted, indicating a promising alternative to n-point statistics in catalogue simulation-based analyses.

astro-ph.CO↗

arXiv2021

Lifting weak lensing degeneracies with a field-based likelihood

We present a field-based approach to the analysis of cosmic shear data to infer jointly cosmological parameters and the dark matter distribution. This forward modelling approach samples the cosmological parameters and the initial matter fluctuations, using a physical gravity model to link the primordial fluctuations to the non-linear matter distribution. Cosmological parameters are sampled and updated consistently through the forward model, varying (1) the initial matter power spectrum, (2) the geometry through the distance-redshift relationship, and (3) the growth of structure and light-cone effects. Our approach extracts more information from the data than methods based on two-point statistics. We find that this field-based approach lifts the strong degeneracy between the cosmological matter density, $Ω_\mathrm{m}$, and the fluctuation amplitude, $σ_8$, providing tight constraints on these parameters from weak lensing data alone. In the simulated four-bin tomographic experiment we consider, the field-based likelihood yields marginal uncertainties on $σ_8$ and $Ω_\mathrm{m}$ that are, respectively, a factor of 3 and 5 smaller than those from a two-point power spectrum analysis applied to the same underlying data.

astro-ph.CO↗

arXiv2021

Bayesian forward modelling of cosmic shear data

We present a Bayesian hierarchical modelling approach to infer the cosmic matter density field, and the lensing and the matter power spectra, from cosmic shear data. This method uses a physical model of cosmic structure formation to infer physically plausible cosmic structures, which accounts for the non-Gaussian features of the gravitationally evolved matter distribution and light-cone effects. We test and validate our framework with realistic simulated shear data, demonstrating that the method recovers the unbiased matter distribution and the correct lensing and matter power spectrum. While the cosmology is fixed in this test, and the method employs a prior power spectrum, we demonstrate that the lensing results are sensitive to the true power spectrum when this differs from the prior. In this case, the density field samples are generated with a power spectrum that deviates from the prior, and the method recovers the true lensing power spectrum. The method also recovers the matter power spectrum across the sky, but as currently implemented, it cannot determine the radial power since isotropy is not imposed. In summary, our method provides physically plausible inference of the dark matter distribution from cosmic shear data, allowing us to extract information beyond the two-point statistics and exploiting the full information content of the cosmological fields.

astro-ph.CO↗

arXiv2020

A hierarchical field-level inference approach to reconstruction from sparse Lyman-$α$ forest data

We address the problem of inferring the three-dimensional matter distribution from a sparse set of one-dimensional quasar absorption spectra of the Lyman-$α$ forest. Using a Bayesian forward modelling approach, we focus on extending the dynamical model to a fully self-consistent hierarchical field-level prediction of redshift-space quasar absorption sightlines. Our field-level approach rests on a recently developed semiclassical analogue to Lagrangian perturbation theory (LPT), which improves over noise problems and interpolation requirements of LPT. It furthermore allows for a manifestly conservative mapping of the optical depth to redshift space. In addition, this new dynamical model naturally introduces a coarse-graining scale, which we exploited to accelerate the Markov chain Monte-Carlo (MCMC) sampler using simulated annealing. By gradually reducing the effective temperature of the forward model, we were able to allow it to first converge on large spatial scales before the sampler became sensitive to the increasingly larger space of smaller scales. We demonstrate the advantages, in terms of speed and noise properties, of this field-level approach over using LPT as a forward model, and, using mock data, we validated its performance to reconstruct three-dimensional primordial perturbations and matter distribution from sparse quasar sightlines.

astro-ph.CO↗

arXiv2019

Inferring high redshift large-scale structure dynamics from the Lyman-alpha forest

One of the major science goals over the coming decade is to test fundamental physics with probes of the cosmic large-scale structure out to high redshift. Here we present a fully Bayesian approach to infer the three-dimensional cosmic matter distribution and its dynamics at $z>2$ from observations of the Lyman-$α$ forest. We demonstrate that the method recovers the unbiased mass distribution and the correct matter power spectrum at all scales. Our method infers the three-dimensional density field from a set of one-dimensional spectra, interpolating the information between the lines of sight. We show that our algorithm provides unbiased mass profiles of clusters, becoming an alternative for estimating cluster masses complementary to weak lensing or X-ray observations. The algorithm employs a Hamiltonian Monte Carlo method to generate realizations of initial and evolved density fields and the three-dimensional large-scale flow, revealing the cosmic dynamics at high redshift. The method correctly handles multi-modal parameter distributions, which allow constraining the physics of the intergalactic medium (IGM) with high accuracy. We performed several tests using realistic simulated quasar spectra to test and validate our method. Our results show that detailed and physically plausible inference of three-dimensional large-scale structures at high redshift has become feasible.

astro-ph.CO↗

arXiv2019

Explicit Bayesian treatment of unknown foreground contaminations in galaxy surveys

The treatment of unknown foreground contaminations will be one of the major challenges for galaxy clustering analyses of coming decadal surveys. These data contaminations introduce erroneous large-scale effects in recovered power spectra and inferred dark matter density fields. In this work, we present an effective solution to this problem in the form of a robust likelihood designed to account for effects due to unknown foreground and target contaminations. Conceptually, this robust likelihood marginalizes over the unknown large-scale contamination amplitudes. We showcase the effectiveness of this novel likelihood via an application to a mock SDSS-III data set subject to dust extinction contamination. In order to illustrate the performance of our proposed likelihood, we infer the underlying dark-matter density field and reconstruct the matter power spectrum, being maximally agnostic about the foregrounds. The results are compared to those of an analysis with a standard Poissonian likelihood, as typically used in modern large-scale structure analyses. While the standard Poissonian analysis yields excessive power for large-scale modes and introduces an overall bias in the power spectrum, our likelihood provides unbiased estimates of the matter power spectrum over the entire range of Fourier modes considered in this work. Further, we demonstrate that our approach accurately accounts for and corrects the effects of unknown foreground contaminations when inferring three-dimensional density fields. Robust likelihood approaches, as presented in this work, will be crucial to control unknown systematic error and maximize the outcome of the decadal surveys.

astro-ph.CO↗

arXiv2017

NIFTy 3 - Numerical Information Field Theory - A Python framework for multicomponent signal inference on HPC clusters

NIFTy, "Numerical Information Field Theory", is a software framework designed to ease the development and implementation of field inference algorithms. Field equations are formulated independently of the underlying spatial geometry allowing the user to focus on the algorithmic design. Under the hood, NIFTy ensures that the discretization of the implemented equations is consistent. This enables the user to prototype an algorithm rapidly in 1D and then apply it to high-dimensional real-world problems. This paper introduces NIFTy 3, a major upgrade to the original NIFTy framework. NIFTy 3 allows the user to run inference algorithms on massively parallel high performance computing clusters without changing the implementation of the field equations. It supports n-dimensional Cartesian spaces, spherical spaces, power spaces, and product spaces as well as transforms to their harmonic counterparts. Furthermore, NIFTy 3 is able to treat non-scalar fields. The functionality and performance of the software package is demonstrated with example code, which implements a real inference algorithm from the realm of information field theory. NIFTy 3 is open-source software available under the GNU General Public License v3 (GPL-3) at https://gitlab.mpcdf.mpg.de/ift/NIFTy/

astro-ph.IM↗