Contributing:

• Primary: Soumadeep Saha, Indian Statistical Institute, Kolkata, India
• Additional:
  • Rahul Shah, Indian Statistical Institute, Kolkata, India
  • Purba Mukherjee, Indian Statistical Institute, Kolkata, India

Paper

This repository contains code supplementing the paper LADDER - Revisiting the Cosmic Distance Ladder with Deep Learning Approaches and Exploring its Applications.

• Rahul Shah, Soumadeep Saha, Purba Mukherjee, Utpal Garain, Supratik Pal

Abstract

We investigate the prospect of reconstructing the ``cosmic distance ladder'' of the Universe using a novel deep learning framework called LADDER - Learning Algorithm for Deep Distance Estimation and Reconstruction. LADDER is trained on the apparent magnitude data from the Pantheon Type Ia supernovae compilation, incorporating the full covariance information among data points, to produce predictions along with corresponding errors. After employing several validation tests with a number of deep learning models, we pick LADDER as the best performing one. We then demonstrate applications of our method in the cosmological context, that include serving as a model-independent tool for consistency checks for other datasets like baryon acoustic oscillations, calibration of high-redshift datasets such as gamma ray bursts, use as a model-independent mock catalog generator for future probes, etc. Our analysis advocates for interesting yet cautious consideration of machine learning applications in these contexts.

Algorithm

Usage:

Note

Requires python version 3.10 or similar.

Warning

CUDA - A CUDA capable system (CUDA version 11.8 or greater) is required to run this.

• Make sure dependencies are installed. If you have python installed run :

pip3 install -r requirements.txt

• After making sure dependencies are installed, run the program with:

python3 main.py --pairs [number] --output [directory name]

• Explanations :

  --pairs [num] this gives the sequence length for LSTM model.
  --output [dir_name] is the output directory to write weights 
  and statistics in (Stored in ./outputs/dir_name)
  

• Example:

python3 main.py -p 25 --output spock_25

For training the spock model on the dataset followed by predictions and writing to ./outputs/spock_25"""

Academic Use

If you use the provided code in the preparation of a manuscript for academic publication, please cite it as (BiBTeX):

@article{Shah:2024slr,
    author = "Shah, Rahul and Saha, Soumadeep and Mukherjee, Purba and Garain, Utpal and Pal, Supratik",
    title = "{LADDER: Revisiting the Cosmic Distance Ladder with Deep Learning Approaches and Exploring Its Applications}",
    eprint = "2401.17029",
    archivePrefix = "arXiv",
    primaryClass = "astro-ph.CO",
    doi = "10.3847/1538-4365/ad5558",
    journal = "Astrophys. J. Suppl. Ser.",
    volume = "273",
    number = "2",
    pages = "27",
    year = "2024"
}

Any commmercial/non-commercial use is permitted with appropriate citation. If a significant portion of the code is used/redistributed it must include a copy of the LICENSE file.

Files:

• README.md -> This file.
• requirements.txt -> List of requirements to run this library.
• utils/
  • dataset.py -> Loads dataset and implements sampling routine.
  • inference.py -> Inference algorithm implementation.
  • train.py -> Train/test loops.
  • utils.py -> Utility functions for saving files, etc.
  • model.py -> Model script.
• data/ -> Data directory.
• main.py -> Main script, runs everything else.
• config.toml -> Config file storing settings, etc.