Deep learning exoplanets detection by combining real and synthetic data

Area of Science:

• Astronomy and Astrophysics
• Machine Learning

Background:

• Exoplanet discovery is crucial, with over 4300 confirmed planets.
• Planetary transit detection is a key method for finding exoplanets.
• Kepler Telescope light curves provide valuable data for exoplanet research.

Purpose of the Study:

• To develop a deep learning system for detecting planetary transits.
• To enhance transit detection using real Kepler Telescope light curves.
• To optimize the use of synthetic data in training deep learning models.

Main Methods:

• A Convolutional Neural Network (CNN) classification model was developed.
• The model was trained on a mix of real and synthetic Kepler light curve data.
• Model performance was validated using unknown, real light curve data.

Main Results:

• The optimal ratio of synthetic data was determined through optimization and sensitivity analysis.
• The trained model demonstrated improved precision, accuracy, and true positive rates.
• Performance was compared favorably against other similar studies.

Conclusions:

• Synthetic data can effectively enhance the training of deep learning models for exoplanet transit detection.
• The developed system shows promise for improving the efficiency and accuracy of exoplanet discovery.
• This approach contributes to advancing the field of exoplanetary science.