A5078513445- Galaxies: Formation, Evolution, Phenomena
- Astronomy and Astrophysical Research
- Cosmology and Gravitation Theories
- Astronomical Observations and Instrumentation
- Seismic Imaging and Inversion Techniques
- Gamma-ray bursts and supernovae
- Seismology and Earthquake Studies
- Dark Matter and Cosmic Phenomena
- Computational Physics and Python Applications
- Statistical and numerical algorithms
- Stellar, planetary, and galactic studies
- Gaussian Processes and Bayesian Inference
- Scientific Research and Discoveries
- Adaptive optics and wavefront sensing
- Astrophysical Phenomena and Observations
- Particle physics theoretical and experimental studies
- Reservoir Engineering and Simulation Methods
- Astrophysics and Cosmic Phenomena
- Radio Astronomy Observations and Technology
- Remote Sensing in Agriculture
- Atmospheric and Environmental Gas Dynamics
- Impact of Light on Environment and Health
- Advanced Physical and Chemical Molecular Interactions
- History and Developments in Astronomy
- Spatial and Panel Data Analysis
University College London
2019-2025
Royal Holloway University of London
2024-2025
Istituto Nazionale di Fisica Nucleare, Laboratori Nazionali di Frascati
2023
University of Cambridge
2019-2022
Heidelberg University
2017-2019
General Relativity and the $\Lambda$CDM framework are currently standard lore constitute concordance paradigm. Nevertheless, long-standing open theoretical issues, as well possible new observational ones arising from explosive development of cosmology last two decades, offer motivation lead a large amount research to be devoted in constructing various extensions modifications. All extended theories scenarios first examined under light consistency, then applied geometrical backgrounds, such...
ABSTRACT We present CosmoPower, a suite of neural cosmological power spectrum emulators providing orders-of-magnitude acceleration for parameter estimation from two-point statistics analyses Large-Scale Structure (LSS) and Cosmic Microwave Background (CMB) surveys. The replace the computation matter CMB spectra Boltzmann codes; thus, they do not need to be re-trained different choices astrophysical nuisance parameters or redshift distributions. emulation error is less than $0.4{{\ \rm per\...
We present a sample-variance-limited measurement of the temperature power spectrum ($TT$) cosmic microwave background using observations $\ensuremath{\sim}1500\text{ }\text{ }{\mathrm{deg}}^{2}$ field made by SPT-3G in 2018. report multifrequency measurements at 95, 150, and 220 GHz covering angular multipole range $750\ensuremath{\le}\ensuremath{\ell}<3000$. combine this $TT$ with published polarization from 2018 observing season update their associated covariance matrix to complete...
Abstract We present a coherent, re-usable python framework building on the CosmoPower emulator code for high-accuracy calculations of cosmological observables with Einstein-Boltzmann codes. For detailed statistical analyses, such codes require high computing power, making parameter space exploration costly, especially beyond-ΛCDM analyses. Machine learning-enabled emulators are becoming an increasingly popular solution to this problem. To enable generation, sharing and use inference, we...
ABSTRACT We present constraints on Horndeski gravity from a combined analysis of cosmic shear, galaxy–galaxy lensing and galaxy clustering $450\, \mathrm{deg}^2$ the Kilo-Degree Survey Galaxy And Mass Assembly survey.The class dark energy/modified models includes majority universally coupled extensions to ΛCDM with one scalar field in addition metric. study functions time that fully describe evolution linear perturbations gravity. Our results are compatible throughout model. By imposing...
We present a tomographic weak lensing analysis of the Kilo Degree Survey Data Release 4 (KiDS-1000), using new pseudo angular power spectrum estimator (pseudo- C ℓ ) under development for ESA Euclid mission. Over 21 million galaxies with shape information are divided into five redshift bins, ranging from 0.1 to 1.2 in photometric redshift. measured pseudo- eight bands multipole range 76 < 1500 auto- and cross-power spectra between bins. A series tests were carried out check systematic...
ABSTRACT We use the emulation framework CosmoPower to construct and publicly release neural network emulators of cosmological observables, including cosmic microwave background (CMB) temperature polarization power spectra, matter spectrum, distance-redshift relation, baryon acoustic oscillation (BAO) redshift-space distortion (RSD) derived parameters. train our on Einstein–Boltzmann calculations obtained with high-precision numerical convergence settings, for a wide range models ΛCDM, wCDM,...
We present the learned harmonic mean estimator with normalizing flows - a robust, scalable and flexible of Bayesian evidence for model comparison. Since is agnostic to sampling strategy simply requires posterior samples, it can be applied compute using any Markov chain Monte Carlo (MCMC) technique, including saved down MCMC chains, or variational inference approach. The was recently introduced, where machine learning techniques were developed learn suitable internal importance target...
This work introduces a novel framework for full-waveform seismic source inversion using simulation-based inference (SBI). Traditional probabilistic approaches often rely on simplifying assumptions about data errors, which we show can lead to inaccurate uncertainty quantification. SBI addresses this limitation by building an empirical model of the errors machine learning models, known as neural density estimators, then be integrated into Bayesian framework. We apply point-source moment tensor...
Summary This paper presents a novel framework for full-waveform seismic source inversion using simulation-based inference (SBI). Traditional probabilistic approaches often rely on simplifying assumptions about data errors, which we show can lead to inaccurate uncertainty quantification. SBI addresses this limitation by learning an empirical relationship between the parameters and data, without making errors. is achieved through use of specialised machine models, known as neural density...
Cosmic shear is one of the primary probes to test gravity with current and future surveys. There are two main techniques analyse a cosmic survey; tomographic method, where correlations between lensing signal in different redshift bins used recover information, 3D approach, full information carried through entire analysis. Here we compare methods, by forecasting cosmological constraints for surveys like Euclid. We extend formalism first time theories beyond standard model, belonging Horndeski...
We present CosmoPower-JAX, a JAX-based implementation of the CosmoPower framework, which accelerates cosmological inference by building neural emulators power spectra. show how, using automatic differentiation, batch evaluation and just-in-time compilation features JAX, running pipeline on graphics processing units (GPUs), parameter estimation can be accelerated orders magnitude with advanced gradient-based sampling techniques. These used to efficiently explore high-dimensional spaces, such...
ABSTRACT We present constraints on the dark scattering model through cosmic shear measurements from Kilo Degree Survey (KiDS-1000), using an accelerated pipeline with novel emulators produced CosmoPower. Our main emulator, for non-linear matter power spectrum, is trained predictions halo reaction framework, previously validated against simulations. Additionally, we include effects of baryonic feedback HMCode2016, whose contribution also emulated. analyse complete set statistics KiDS-1000,...
Future observations of the large-scale structure have potential to investigate cosmological models with a high degree complexity, including properties gravity on large scales, presence complicated dark energy component, and addition neutrinos changing structures small scales. Here we study Horndeski theories gravity, most general minimally coupled scalar-tensor second order. While background evolution can be described by an effective equation state, perturbations are characterised four free...
Abstract Comparison of appropriate models to describe observational data is a fundamental task science. The Bayesian model evidence, or marginal likelihood, computationally challenging, yet crucial, quantity estimate perform comparison. We introduce methodology compute the evidence in simulation-based inference (SBI) scenarios (often called likelihood-free inference). In particular, we leverage recently proposed learned harmonic mean estimator and exploit fact that it decoupled from method...
ABSTRACT We derive constraints on a coupled quintessence model with pure momentum exchange from the public ∼1000 deg2 cosmic shear measurements Kilo-Degree Survey and Planck 2018 microwave background data. compare this Lambda cold dark matter find similar χ2 log-evidence values. accelerate parameter estimation by sourcing cosmological power spectra neural network emulator CosmoPower. highlight necessity of such emulator-based approaches to reduce computational runtime future analyses,...
class_sz is a versatile, robust and efficient code, in C Python, optimized to compute theoretical predictions for wide range of observables relevant cross-survey science the Stage IV era. The code public at https://github.com/CLASS-SZ/class_sz along with series tutorial notebooks ( https://github.com/CLASS-SZ/notebooks ). It will be presented full detail paper II. Here we give brief overview key features usage.
We advocate for a new paradigm of cosmological likelihood-based inference, leveraging recent developments in machine learning and its underlying technology, to accelerate Bayesian inference high-dimensional settings. Specifically, we combine (i) emulation, where model is trained mimic observables, e.g. CosmoPower-JAX; (ii) differentiable probabilistic programming, JAX NumPyro, respectively; (iii) scalable Markov chain Monte Carlo (MCMC) sampling techniques that exploit gradients, Hamiltonian...
The Fisher matrix is a widely used tool to forecast the performance of future experiments and approximate likelihood large data sets. Most forecasts for cosmological parameters in galaxy clustering studies rely on approach large-scale like DES, Euclid, or SKA. Here we improve upon standard method by taking into account three effects: finite window function, correlation between redshift bins, uncertainty bin redshift. first two effects are negligible only limit infinite surveys. third effect,...
Modelling nonlinear structure formation is essential for current and forthcoming cosmic shear experiments. We combine the halo model reaction formalism, implemented in REACT code, with COSMOPOWER machine learning emulation platform, to develop publicly release REACTEMU-FR (https://github.com/cosmopower-organization/reactemu-fr), a fast accurate matter power spectrum emulator $f(R)$ gravity massive neutrinos. Coupled state-of-the-art baryon feedback BCEMU, we use produce Markov Chain Monte...
We present constraints on the Dark Scattering model through cosmic shear measurements from Kilo Degree Survey (KiDS-1000), using an accelerated pipeline with novel emulators produced $\tt{CosmoPower}$. Our main emulator, for non-linear matter power spectrum, is trained predictions halo reaction framework, previously validated against simulations. Additionally, we include effects of baryonic feedback $\tt{HMcode2016}$, whose contribution also emulated. analyse complete set statistics...
We use the emulation framework CosmoPower to construct and publicly release neural network emulators of cosmological observables, including Cosmic Microwave Background (CMB) temperature polarization power spectra, matter spectrum, distance-redshift relation, baryon acoustic oscillation (BAO) redshift-space distortion (RSD) derived parameters. train our on Einstein-Boltzmann calculations obtained with high-precision numerical convergence settings, for a wide range models $\Lambda$CDM, $w$CDM,...
Cosmic shear---the weak gravitational lensing effect generated by fluctuations of the tidal fields large-scale structure---is one most promising tools for current and future cosmological analyses. The spherical-Bessel decomposition cosmic shear field (3D shear) is way to maximize amount redshift information in a analysis therefore provides powerful tool investigate particular growth structure that crucial dark energy studies. However, computation simulated 3D covariance matrices presents...
Abstract. We present a series of new open-source deep-learning algorithms to accelerate Bayesian full-waveform point source inversion microseismic events. Inferring the joint posterior probability distribution moment tensor components and location is key for rigorous uncertainty quantification. However, inference process requires forward modelling traces each set parameters explored by sampling algorithm, which makes very computationally intensive. In this paper we focus on accelerating...