Skip to content

Commit

Permalink
scripts for the paper
Browse files Browse the repository at this point in the history
  • Loading branch information
@jsv1206
jsv1206 committed Jul 23, 2024
1 parent e50dae2 commit b09ae35
Show file tree
Hide file tree
Showing 15 changed files with 5,899 additions and 142 deletions.
Binary file added .DS_Store
View file
Binary file not shown.
236 changes: 94 additions & 142 deletions README.md
View file
Original file line number Diff line number Diff line change
@@ -1,152 +1,104 @@
This template is designed to give a framework for public distributions of "science" projects.
It is a guideline, showing the minimum things recommended to include with your public repository,
to make your results easily replicable.
It is not exhaustive by any means, nor is everything here strictly required in all cases!
Please consider this as a loose list of things considered "nice to have", and as reference material above all.

# DeepSkies Science Repo Template
Include status links to different outside resources, such as build info, paper info, license, etc.
You can select your license from the [choose a license page.](https://choosealicense.com/licenses/), and then change the name of the license in the badge and link included.
For workflows, change the name of the repo listed in the img.shields link to point to your repo and workflows.

[![status](https://img.shields.io/badge/arXiv-000.000-red)](arxiv link if applicable)
[![status](https://img.shields.io/badge/PyPi-0.0.0.0-blue)](pypi link if applicable)
[![status](https://img.shields.io/badge/License-MIT-lightgrey)](MIT or Apache 2.0 or another requires link changed)
![GitHub Workflow Status](https://img.shields.io/github/workflow/status/owner/repo/build-repo)
![GitHub Workflow Status](https://img.shields.io/github/workflow/status/owner/repo/test-repo?label=test)

Your overview should contain a brief summary of the project, and figures and examples showing input and output.

## Installation
Information about install.
We recommend publishing to pypi using a poetry package management system (described below) but we also provide instructions for using python virtual environments and showyourwork with conda integration.
We present **DeepCosmoSLIDE**: a **S**trong **L**ens based **I**nference for **D**ark **E**nergy. We use Simulation Based Inference (SBI) with Neural Ratio Estimation (NRE) to constrain Dark Energy parameter from a population of strong galaxy-galaxy lenses.

Example of what your installation instructions should look like:

To install with pip:
> pip install git+https://github.com/DeepSkies/science_template.git
>
This will set up a virtual environment, which can b e run with on mac or linux
> source venv/bin/activate
>
Or on windows with
> venv\Scripts\activate.bat
## Introduction

Verify installation is functional is all tests are passing
> pytest
Strong gravitational lensing offers crucial insights into cosmology, but traditional Monte Carlo methods for cosmological inference are computationally prohibitive and inadequate for processing the thousands of lenses anticipated from future cosmic surveys. New tools for inference, such as SBI using NRE, address this challenge effectively. NRE is a classifier neural network to differentiate between two probability distributions:

Additionally, include how to install from source (via git clone) and associated setup.
$(x,w) \sim\ p(x,w)$ with class label y=1

### poetry
Poetry is our recommended method of handling a package environment as publishing and building is handled by a toml file that handles all possibly conflicting dependencies.
Full docs can be found [here](https://python-poetry.org/docs/basic-usage/).

Install instructions:

Add poetry to your python install
> pip install poetry
Install the pyproject file
> poetry install
To add another package to your environment
> poetry add (package name)
To run within your environment
>poetry run (file).py
If you wish to start from scratch:
> pip install poetry
> poetry init
### virtual environment
At the bare minimum, project dependencies must be contained and strictly defined and shared for replication purposes.
The easiest way to do this is to use a python virtual environment.
Full instructions are found [here.](https://docs.python.org/3/library/venv.html)

To initialize an environment:
> python3 -m venv /path/to/env
>
To activate it:
Linux and Mac:
> source venv/bin/activate
>
Windows:
> venv\Scripts\activate.bat
And use pip as normal to install packages.

In order to produce a file to share with your version of dependencies, produce a requirements.txt.
This can later be installed in full to a new system using `pip install -r requirements.txt`.
Note that this does not manage any versioning conflicts and can very quickly become depreciated.
> pip freeze >requirements.txt
### show your work with conda
We also supply a ["show your work"](https://github.com/showyourwork/showyourwork) workflow to use with a conda venv which can compile the example tex file in `DeepTemplate-Science/src/tex/ms.tex`

To execute this workflow:
>showyourwork build
This will build your project and install the conda venv associated with the project (or just compile the document if you haven't been using it) and output the document as a pdf.
If you would like to integrate with overleaf to push your work remotely, you can do that by adding the following lines to your showyourwork.yml file:
>
> overleaf:
>
> id: URL identifying your project
> push:
> - src/tex/figures
> - src/tex/output
> pull:
> - src/tex/ms.tex
> - src/tex/bib.bib
And adding the system variables `$OVERLEAF_EMAIL` and `$OVERLEAF_PASSWORD` with your credentials.
To do this, use a bash terminal to input the command `export OVERLEAF_EMAIL='[email protected]`, and do the same for your password.
To verify these are set correctly, run `echo $OVERLEAF_EMAIL`and `echo $OVERLEAF_PASSWORD`.
To complete this setup, run `showyourwork build` as if you were compiling a project.
The above snippet of the yaml file will then push anything in the `src/tex/figures` and `src/tex/output` folders to the remote, under the `images` folder.

The existing yaml file is set up to modify the [template project](*https://www.overleaf.com/read/fsjwntpjmdzw).
The differences in the ID in the template and the url you'll see is due to the fact that only project owners have access to that ID (even if I want to share).
This limits the person who can build the project to the person that owns the overleaf page, at least until Latex sets up token authentication.
The workaround for this is account sharing, but this is not recommended.

For more information please see the [showyourwork page on the topic](https://show-your.work/en/latest/overleaf/).
$(x,w) \sim\ p(x)p(w)$ with class label y=0

where $x$ is the strong lens image and $w$ is the dark energy equation-of-state parameter that generated the image.

By training a machine learning model on simulated data of strong lenses, we can learn the likelihood-to-evidence ratio $\frac{p(x|w)}{p(x)}$. This is used for robust inference of the posterior $p(w|\{x_{0}\})$ from a population of strong lens images $\{x_{0}\}$.

$\textbf{Analysis Workflow}$

The following figure summarizes the workflow of the analysis.

The strong lens images are genereated using a simulator where the parameters are sampled from a prior distribution. The training data for the NRE network (classifier) includes the image and the parameter of interest. The network outputs the likelihood-to-evidence ratio. The trained model is implemented on the observations to estimate the posterior.

![Workflow](./figures/SBI_NRE_workflow.png)

## Getting started

### Data

The data used for this analysis can be found on Zenodo (link will be provided shortly).

The images are generated using [Deeplenstronomy](https://github.com/deepskies/deeplenstronomy) package.

This data can be generated using the yaml files in `/deeplenstornomy_templates` as inputs to $\texttt{Deeplenstronomy}$.

The simulation outputs the data into a folder which includes images (`/CONFIGURATION_1_images.npy`) and the metadata (`/CONFIGURATION_1_metadata.csv`) assocated with the image generation.

$\textbf{Training data}$
We train, validate, and test the NRE model on simulated data of 1M strong lens images.

$\textbf{Test data for population-level analysis}$
We generate three datasets of 3000 images each by fixing w = -1.2, -1.0, and -0.8 respectively.


<!-- #### Architecture
The network is a ResNet architecture to classify between two classes. We train the network by minimizing the Brinay Cross Entropy (BCE) loss function by dynamically adjusting the learning rate when the vali-
dation loss is plateaued by a decay factor of 0.1 starting
from $1e^{−2}$ to $1e^{−6}$, if the validation loss does not improve
over five epochs. We include an option for early stopping if the validation loss does not improve over 20 epochs. The model is trained on NVIDIA A100 GPU for 71 epochs with a typical training time of 25
minutes. -->

### Notebooks for analysis

The notebooks in `/notebooks` can be run to reproduce the results of the paper. The python scripts are currently being developed to be run from the terminal.

$\textbf{Model Training}$

`train_model.ipynb`
This notebook includes reading in the data, preprocessing of the images,
and training the model.
Three models with random weight initializations (`seed` input to the model) are run in our analysis to check robustness. One of the models `working_model_1M-2-034_seed128_v2.keras` is available on Zonodo (link will be provided shortly).

This trained model can be directly loaded (without having to re-train the model) using

`model = tf.keras.models.load_model(model_name)`

$\textbf{Model Evaluation}$

`compare_random_seeds.ipynb` code is for checking the performance of the model on test data of 2000 images.
The code includes plotting the Receiver Operating Curve (ROC) for the three models. It also includes calculating and plotting the analytical posteriors of a few randomly selected strong lenses.

`plot_image_posterior.ipynb` code is for plotting the training data and show the correlation between the Einstein radius and $w$. This code also plots the image of strong lens from the training data and the corresponing analytical posterior.

`plot_residuals` is to plot the predicted mean $w$ of the analytical posterior with 1 $\sigma$ error bar Vs the true $w$ of the 2000 test images. We also compute the posterior coverage plot to check the model uncertainity.

$\textbf{Population-level Analysis}$

`NRE_varyastro_w12.ipynb`, `NRE_varyastro_w1.ipynb`, and `NRE_varyastro_w08.ipynb` include functions to compute the joint posterior from 3000 images with $w$ fixed to -1.2, -1.0, and -0.8 respectively. We show both MCMC and analytical methods to calculate the posterior.

`Compare_mcmc_analytical.ipynb` compares the posteriors from the MCMC and analytical methods.


### Authors

Sreevani Jarugula

### References

If you use this code, please cite our paper (Link to be posted shortly)















## Quickstart
Description of the immediate steps to replicate your results, pointing to a script with cli execution.
You can also point to a notebook if your results are highly visual and showing plots in line with code is desired.

Example:

To run full model training:
> python3 train.py --data /path/to/data/folder
To evaluate a single ""data format of choice""
> python3 eval.py --data /path/to/data
## Documentation
Please include any further information needed to understand your work.
This can include an explanation of different notebooks, basic code diagrams, conceptual explanations, etc.
If you have a folder of documentation, summarize it here and point to it.

## Citation
Include a link to your bibtex citation for others to use.

```
@article{key ,
author = {You :D},
title = {title},
journal = {journal},
volume = {v},
year = {20XX},
number = {X},
pages = {XX--XX}
}
```

## Acknowledgement
Include any acknowledgements for research groups, important collaborators not listed as a contributor, institutions, etc.
Binary file added deeplenstronomy_templates/.DS_Store
View file
Binary file not shown.
134 changes: 134 additions & 0 deletions deeplenstronomy_templates/test_8_parameter_w0cosmo_fixzv_fixw0-08.yml
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,134 @@
DATASET:
NAME: w0_8param_fixzv_test_fixw0-08_3000
PARAMETERS:
SIZE: 3000 # number of images in the full datase.
OUTDIR: w0_8param_fixzv_test_fixw0-08_3000 # will be created on your system if your request to save images
SEED: 42

COSMOLOGY:
NAME: 'wCDM'
PARAMETERS:
H0: 70
Om0: 0.3
Ode0: 0.7
w0: -0.8
IMAGE:
PARAMETERS:
exposure_time:
DISTRIBUTION:
NAME: des_exposure_time
PARAMETERS: None
numPix: 32
pixel_scale: 0.263
psf_type: 'GAUSSIAN'
read_noise: 7
ccd_gain:
DISTRIBUTION:
NAME: des_ccd_gain
PARAMETERS: None

SURVEY:
PARAMETERS:
BANDS: g
seeing: 0.9
magnitude_zero_point: 30.0
sky_brightness: 30.0
num_exposures: 10

SPECIES:
GALAXY_1:
NAME: LENS
LIGHT_PROFILE_1:
NAME: SERSIC_ELLIPSE
PARAMETERS:
magnitude: 100
center_x: 0
center_y: 0
R_sersic: 1
n_sersic: 4
e1: 0
e2: 0.5

MASS_PROFILE_1:
NAME: SIE
PARAMETERS:
sigma_v: 200
e1:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: -0.1
maximum: 0.1
e2:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: -0.1
maximum: 0.1
center_x: 0.0
center_y: 0.0


GALAXY_2:
NAME: SOURCE
LIGHT_PROFILE_1:
NAME: SERSIC_ELLIPSE
PARAMETERS:
magnitude:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: 19
maximum: 24
center_x: 0.0
center_y: 0.0
R_sersic:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: 0.1
maximum: 3
n_sersic:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: 0.5
maximum: 8
e1:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: -0.1
maximum: 0.1
e2:
DISTRIBUTION:
NAME: uniform
PARAMETERS:
minimum: -0.1
maximum: 0.1
MASS_PROFILE_1:
NAME: SIE
PARAMETERS:
theta_E: 2.0
e1: 0.1
e2: -0.1
center_x: 0.0
center_y: 0.0


GEOMETRY:
CONFIGURATION_1:
NAME: GALAXYGALAXY
FRACTION: 1
PLANE_1:
OBJECT_1: LENS
PARAMETERS:
REDSHIFT: 0.1
PLANE_2:
OBJECT_1: SOURCE
PARAMETERS:
REDSHIFT: 2.0




Loading

0 comments on commit b09ae35

Please sign in to comment.