# **Fri 13th May**

**To do**

Short term (next few days):
* Store arrays of initial and final density fields for \Omega_{cdm} values 0 to 1; growth factors for each field
* Use PyTorch to determine the growth factor by fitting a curve taking \Omega_{cdm} as the input.
* Check how the growth factor matches with analytic formula
* Summary stats - histograms, power spectra,

Medium term (next week or so):
* Attempt to incorporate translation and rotation symmetries into model
* Apply techniques to data with Non-Gaussian clustering.

Reading:
* Normalizing flows for random fields in cosmology (2021) https://arxiv.org/pdf/2105.12024.pdf
* Flow-based likelihoods for non-Gaussian inference https://arxiv.org/pdf/2007.05535.pdf
* FlowPM: Distributed TensorFlow Implementation of the FastPM Cosmological N-body Solver https://arxiv.org/abs/2010.11847
* Inference and De-Noising of Non-Gaussian Particle Distribution Functions: A Generative Modeling Approach https://arxiv.org/pdf/2110.02153.pdf

# **Thursday 19th May**

**Task 1 - Computing the Growth Factor**
explanation of the growth factor https://academic.oup.com/mnras/article/322/2/419/963632

growth factor:
g = D/a; where D is amplitude of growing mode (Fourier transforms?) and a is the cosmic scale-factor

Formula for the growth factor is given by:


Therefore, the growth factor depends on the matter vs DE ratio and redshift.

Approximation given in the article:


I'm not 100% sure but I think this is for the z = 0 case specifically.

CSF:
The cosmic scale factor, a, depends on the matter content of the universe (i.e. omega_cdm and omega_de).

Subtasks:
* Is there a quick way to perform the integral to compute the growth factor at a particular redshift?
* Find out why cosmo.growth_factor(z=0) doesn't want to work?
* Using an initial density field, find out how to use the growth factor ratio to get to the final density field.