This Github organization puts together several codes, whose aim is to emulate cosmological observables as predicted by Einsten-Boltzmann solvers and Perturbation Theory codes. The main programming language employed in these repositories is Julia, but we are working on Python versions that are based on JAX.
Actually, the observables we emulates are:
- CMB angular Power Spectrum, with
Capse.jlandjaxcapse - Galaxy Clustering Power Spectrum multipoles based on EFT with
Effort.jlandjaxeffort - Linear and Nonlinear Matter Power Spectra with
Mapse.jl - BAO correlation function with
Bora.jl
We also provide a package, EmulatorsTrainer.jl, that has utilities to create training datasets, train emulators, and validate their performance.
Currently, we employ two different neural network backends for the Julia emulators:
SimpleChains.jl, a high-performance framework tailored for small NNs running on a CPULux.jl, which is fully GPU compatible
Although the former is (in general) faster for our applications, the latter opens to the possibility of using ensamble samplers, such as MicroCanonical Hamiltonian MonteCarlo, that can easily run on a GPU. For the JAX emulators, we employ flax as the NN-backend.
Our emulators are differentiable, i.e. we can use automatic (also dubbed algorithmic) differentiation in order to evaluate derivatives. This enable for gradient-based methods, such as the minimization L-BFGS algorithm (as implemented in Optim.jl) or the Hamiltonian MonteCarlo inference algorithm (as implemented in Turing.jl). The same is true for our JAX-based emulators, which we explicitely check can be differentiated end-to-end.
Our codes have been officially released in the following publications:
- Bonici, Bianchini, and Ruiz-Zapatero, Capse.jl: efficient and auto-differentiable CMB power spectra emulation
- Bonici, D'Amico, Bel, and Carbone, Effort.jl: a fast and differentiable emulator for the Effective Field Theory of the Large Scale Structure of the Universe
The codes previously listed are used in the following publications:
- Crespi, Percival, Krolewski, Bonici, et al., Baryon fraction from the BAO amplitude: a consistent approach to parameterizing perturbation growth
- Krolewski, Crespi, Percival, Bonici, et al., A measurement of H0 from DESI DR1 using energy densities
- Crespi, Bonici, Loureiro, et al., Flinch: A Differentiable Framework for Field-Level Inference of Cosmological parameters from curved sky data
- Morawetz, Zhang, Bonici, Percival, Crespi, et al., Frequentist Cosmological Constraints from Full-Shape Clustering Measurements in DESI DR1
- Zhang, Bonici, Rocher, Percival, de Mattia, et al., Enhancing DESI DR1 Full-Shape analyses using HOD-informed priors
- Baleato Lizancos, Seljak, Karamanis, Bonici, Ferraro, Selecting samples of galaxies with fewer Fingers-of-God
- Paradiso, Bonici, Chen, Percival, D'Amico, Zhang, and McGee, Reducing nuisance prior sensitivity via non-linear reparameterization, with application to EFT analyses of large-scale structure
- SPT Collaboration, Cosmology From CMB Lensing and Delensed EE Power Spectra Using 2019-2020 SPT-3G Polarization Data
- Zhang, Bonici, D'Amico, Paradiso, and Percival, HOD-informed prior for EFT-based full-shape analyses of LSS
