Temporal Convolutional Networks for Exoplanet Transit Timing Variations

Objective

Develop a machine learning approach using Temporal Convolutional Networks (TCNs) to detect and analyze transit timing variations (TTVs) in multi-planet systems. This method aims to reveal hidden planets or provide deeper insights into planetary dynamics.

Table of Contents

1. Installation
   • Clone the Repository
   • Create a Virtual Environment
   • Install Dependencies
2. Data Collection
   • Kepler
   • K2
   • TESS
3. Data Preprocessing
   • [Step 1](#Step 1)
   • [Step 2](#Step 2)
   • Download Link
4. Running the Model

Installation

Clone the Repository

To get started, clone the repository to your local machine using the following command:

git clone https://github.com/your-username/your-repository.git

Navigate into the project directory:

cd your-repository

Create a Virtual Environment

Create a virtual environment to manage dependencies:

python -m venv env

Activate the virtual environment:

• On Windows:

  .\env\Scripts\activate

• On macOS and Linux:

  source env/bin/activate

Install Dependencies

Install the required dependencies using pip.

Requirements for data + data preprocessing

pip install -r requirements/data-requirements.txt

Requirements for ml model

pip install -r requirements/model-requirements.txt

Data Collection

Use the already provided csv files from the missions and run the respective python scripts to download the data.

cd data_collection

Kepler

Use the Kepler Objects of Interest 2024-07-23.csv to download the data from the Kepler mission.

To download the Kepler data, run the following command:

python3  kepler.py 

K2

Use the K2 Planets July 23.csv to download the data from the K2 mission.

To download the Kepler data, run the following command:

python3  k2.py 

TESS

Use the TESS Project Candidates 2024-07-23.csv to download the data from the K2 mission.

To download the Kepler data, run the following command:

python3  tess.py 

Data Preprocessing

Cleaning the light curve and taking only information that we need. This includes cleaning, transforming, and splitting the data.

cd preprocessing

Step 1

Run the general data preprocessing script:

python preprocess.py

Step 2

Run the script to prepare the data to be feed into the ml model:

python preprocess_ml.py

Download Link

Download the preprocessed data (Around 5.5 GB)

Running the Model

Run the custom TCN model to detect TTVs in the light curves.

cd ml

Train + Evaluate the Model

Change the location of the data in the actual_model.py file to where the data is stored.

data = np.load("../ml_data/ttv-dataset/ttv_detection_data.npz") 

Run the TCN model

python actual_model.py

Research Papers

1. "Transit Timing Variations as a Method for Detecting Exoplanets":

   • Holman, M. J., & Murray, N. W. (2005). Science, 307(5713), 1288-1291.
   • This foundational paper discusses the TTV method and its potential for detecting exoplanets.

2. "Application of Convolutional Neural Networks to Exoplanet Detection in Light Curves":

   • Shallue, C. J., & Vanderburg, A. (2018). The Astronomical Journal, 155(2), 94.
   • Explores the use of convolutional neural networks for detecting exoplanets, providing a basis for adapting CNN techniques to TCNs for TTV analysis.

3. "Neural Network Approaches to TTV Detection":

   • Pearson, K. A., Palafox, L., & Griffith, C. A. (2018). Monthly Notices of the Royal Astronomical Society, 474(4), 4782-4796.
   • Investigates the application of neural networks for transit timing variation detection.

4. "Machine Learning for Exoplanet Detection and Characterization":

   • Armstrong, D. J., Pollacco, D., & Santerne, A. (2017). Monthly Notices of the Royal Astronomical Society, 465(3), 2634-2654.
   • Reviews various machine learning methods used in exoplanet detection, including TTV analysis.