Scientific Domain Knowledge Improves Exoplanet Transit Classification with Deep Learning
Problem
Space missions like Kepler and TESS produce large datasets of transit signals that must be efficiently and systematically classified as exoplanets or false positives.
Why it matters
Efficient automated vetting is needed to find planets, including rocky habitable-zone candidates, in huge datasets.
What the AI did
A deep learning model (building on Shallue & Vanderburg) was enhanced with scientific domain knowledge—centroid time-series and stellar parameters—to automatically classify Kepler transit signals.
Results
97.5% accuracy and 98.0% average precision, up from 95.8% and 95.5%; 15–20% recall gains for lowest signal-to-noise transits.
Impact
Improved automated exoplanet vetting, notably boosting recall for low-SNR transits corresponding to rocky habitable-zone planets.
Record
| Venue | The Astrophysical Journal Letters |
|---|---|
| Published | 2018 |
| Work type | article |
| Authors | Megan Ansdell, Yani Ioannou, Hugh P. Osborn, Michele Sasdelli, (2018 NASA Frontier Development Lab Exoplanet Team), Jeffrey C. Smith, Douglas Caldwell, Jon M. Jenkins et al. |
| Citations | 77 (OpenAlex, at import time) |
| References | 24 |
| Model or method | Deep learning model per Shallue & Vanderburg, extended with domain knowledge (architecture not named) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3847/2041-8213/aaf23b |
| Open access | bronze |
| Keywords | exoplanets, transit classification, deep learning, Kepler, false positives, habitable zone |
Importance rating
significant — thousands to millions affected
Automated vetting of large Kepler/TESS datasets to find planets, including rocky habitable-zone candidates; a genuinely new capability for a foundational science, no lives/industry scale.
More in Physics & Astronomy
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | Magnetic control of tokamak plasmas through deep reinforcement learning | 2022 | 739 |
7
|
| 2 | Solving the quantum many-body problem with artificial neural networks | 2017 | 2,205 |
7
|
| 3 | Learning high-accuracy error decoding for quantum processors | 2024 | 90 |
7
|
| 4 | Optimizing Quantum Error Correction Codes with Reinforcement Learning | 2019 | 158 |
7
|
| 5 | Machine-learning-assisted correction of correlated qubit errors in a topological code | 2018 | 141 |
7
|