Discovery of the Smallest Ever Ultra-Short-Period Planet Using Novel Phase Folding Detection System Parallelized on a Cheap GPU
ISEF · 2023 Physics and Astronomy First Award
Overview
The longstanding question of how ultra-short-period (USP) planets form and survive so close to their host stars signifies a gap in our understanding of planetary formation. With only 127 USPs confirmed from the NASA Archive, the scarcity of samples hinders investigation. Searching for USPs which have avoided detection requires significantly higher sensitivity and faster computation than existing transit detection methods. I designed ExoScout, an exoplanet detection system comprising a new phase folding algorithm and a Convolutional Neural Network (CNN) detector. My phase folding algorithm enables parallelization for the first time to efficiently handle the computational workload, which when run on a cheap Graphics Processing Unit (GPU), increases speed by 120 times over the traditional Box-fitting Least Squares method. I trained the CNN to identify weak signals in light curves using a simulated dataset to address lack of samples, achieving 97% validation accuracy and recovering all known USPs from a blind search. With ExoScout, I report the discovery of the smallest USP ever detected and a rare USP found in adversely hot conditions around a high temperature F-dwarf, only the eleventh of its kind from the NASA Archive. ExoScout’s increased efficiency and sensitivity allows new discoveries from large-scale datasets from TESS, James Webb, and Earth 2.0 missions, increasing sample size to advance our study of planetary formation. The innovation of this GPU-parallelized folding algorithm run on cheap hardware replaces expensive supercomputers to make research more accessible and can be applied for high-precision periodic signal detection in many fields.
Awards (2)
- First Award of $5,000 $5,000
- George D. Yancopoulos Innovator Award
Competition history
- ISEF 2023
Resources
Related projects
ISEF · 2024
Discovery of Two Potentially Habitable Super-Earths! StealthPlanetFinder: Innovative & Computationally Efficient Algorithms to Detect Exoplanets Overcoming Low Signal-to-Noise Ratio
ISEF · 2021
A Mathematically Driven Physical Analysis into Exoplanet Detection Confirmation Surveys Using Bayesian Inferential Statistics, Machine Learning/Linear Algebra Sklearn and TensorFlow Techniques, the BATMAN Python Programming Library, 3D Printing Techniques, and a Custom-made Python Processing Pipeline for Image Analysis and Lightcurve Detection
ISEF · 2018
Challenging Limitations: Using Deep Learning, Time Series Analysis, and Statistical Methods for Noise Reduction to Develop an Innovative Approach to Exoplanet Candidate Detection Using Earth-Based Telescopes
ISEF · 2024
Discovery of New Extragalactic Planet Candidates: A Novel End-to-End Machine Learning Pipeline for Efficient Transit Detection in the X-ray Spectrum
Closest projects by meaning, across every fair and year in the corpus.
Source: Regeneron International Science and Engineering Fair