marcobonici/cosmocentral.jl

A Julia code for cosmological analysis.

Marco Bonici

PostDoctoral Researcher at Waterloo Centre for Astrophysics

Cosmology & Bayesian Inference

I am a PostDoctoral Researcher at Waterloo Center for Astrophysics in Canada. I am a cosmologist and my research lies in the context of the large-scale-structure of the Universe. Currently my research is focused on methods to accelerate the analysis of cosmological datasets, employing surrogate models and gradient based methods (and most of this work is performed using the Julia programming language).

I am involved in the Euclid mission of the European Space Agency. Euclid will map the galaxy distribution in about a third of the sky. I am mainly involved in the activity of the Galaxy Clustering Science Working Group (GC-SGW).

I am leading a GitHub organization, CosmologicalEmulators, which contains several codes to emulate cosmological observables. Right now we are focusing on Julia-based emulators, although we plan to add support to JAX emulators.

Contact

[email protected]

cosmocentral.jl's Issues

Add global const speed of light

CosmoCentral.jl/src/AngularCoefficients.jl

Lines 57 to 58 in c7aeba8

	c_0 = 2.99792458e5 #TODO: find a package containing the exact value of
	#physical constants involved in calculations

declare a global constant c0, with the correct value once for all.

Nuisance parameters

Add the nuisance parameters (galaxy bias, intrinsic alignment)

Source Density

Introduce a source density struct

Load Cosmological Grid

The Grid in the Angular coefficients evaluation should be taken from the hdf5.

Unequal spacing integration

Altough in the standard approach the sampling in regular, this will not be the case for the Beyond-limber case. In this situation, it is better to implement the non even trapezoidal or the simpson rulehttps://github.com/uwhpsc-2016/homework3/issues/7.

Intrinsic Alignments

Add the intrinsic alignments to the lensing weight function

Density optimization

In this moment the evaluation of the density normalization and density grid is quite expensive (compared to other parts of the code).
In particular, the quadgk integration routines have been set with a very restrictive requirement. A first check showed that, lowering the bar from rtol=1e-12 to 1e-6, the pad of the Cl's is at about the 1e-10 level. Make a more profound check, using Fisher forecasts marginalized errors.

New model Galaxy Bias

Add a new model for the galaxy bias, the one presented in this paper.

Weak Lensing Weight Function

Add the Weak Lensing Weight Function

flat-non flat cosmologies

In this moment, to fix a problem, flat and non flat Cosmologies have been separated. However, we could keep a single w0waCDM cosmology if the forecasting part was written in a more clever way.

Homogeneous case

CosmoCentral.jl/src/CosmologicalStructures.jl

Line 42 in 7285f07

w0::T = -1.

Here the radiation and curvature density have a lower case letter, while other are upper case. Homogeneize them.

Homogeneous case

CosmoCentral.jl/src/CosmologicalStructures.jl

Line 42 in 7285f07

w0::T = -1.

Here the radiation and curvature density have a lower case letter, while other are upper case. Homogeneize them.

Simpson Integration

Add Simspon integration by means of a weight array inside the for loop evaluation

Change Lensing Efficiency Definition

Change the definition of the lensing efficiency. Actually there are two different definitions, one based on the Euclid IST:F Forecast recipe, one on the Beyond Limber paper. Please, choose only one definition and implement it.

Check avx safety

Add a check in the cl evaluation when the evaluation is performed using avx (a simple while loop should do the job).

Repo too big

The repository is way too big.
This is due to the dimension of the documentation and probably due to the presence of some binary files.
In order to reduce the repo dimension, we should:

1. Fix the cleanup. This should considerably reduce the weigh of the documentation.
2. Remove the binary files.

Galaxy Clustering Weight function

Add the Galaxy Clustering Weight Function

ComputeBias

The function ComputeBiasOverGrid should not require "Bias" as argument, since this is already included in the GalaxyClustering Weight function.

Homogeneous case

CosmoCentral.jl/src/CosmologicalStructures.jl

Line 42 in 7285f07

w0::T = -1.

Here the radiation and curvature density have a lower case letter, while other are upper case. Homogeneize them.

Multiple dispatch

Actually we are using multiple dispatch on struct such as w0waCDMCosmologyStruct, which are of the kind w0waCDMCosmology and is contained in AbstractCosmology.
Can we directly use multiple dispatch without using the intermediate w0waCDMCosmology? Please check.

Docker Images and CI

In order to use calssy in the tests, we need to use a Docker image in CI.
Here is the link to a site which explains what to do.
link

Figure of bias

Add the possibility to compute the shift on cosmological parameter determination due to error on Cl prediction

Equation 6 of this paper

Introduce a convolved density struct

Simpson Integration Lensing Efficiency

Due to the complicated nature of the Lensing Efficiency integration, we must define a matrix weight function for the lensing efficiency integral evaluation

Restrict type

CosmoCentral.jl/src/Density.jl

Line 52 in cd98c3e

InstrumentResponse::InstrumentResponse) where {T, I}

i must be a Int. Modify this accordingly.