Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ExoMiner++: Enhanced Transit Classification and a New Vetting Catalog for 2-Minute TESS Data

Hamed Valizadegan, Miguel J. S. Martinho, Jon M. Jenkins, Joseph D. Twicken, Douglas A. Caldwell, Patrick Maynard, Hongbo Wei, William Zhong, Charles Yates, Sam Donald, Karen A. Collins, David Latham, Khalid Barkaoui, Michael L. Calkins, Kylee Carden, Nikita Chazov, Gilbert A. Esquerdo, Tristan Guillot, Vadim Krushinsky, Grzegorz Nowak, Benjamin V. Rackham, Amaury Triaud, Richard P. Schwarz, Denise Stephens, Chris Stockdale, Cristilyn N. Watkins, Francis P. Wilkin

TL;DR

ExoMiner++ advances transit signal classification for TESS 2-minute data by incorporating new diagnostic inputs (e.g., periodogram, flux trend, difference image, momentum dumps) and leveraging multi-source learning with Kepler data. The model architecture integrates multiple input branches and uses Savitzky-Golay detrending, uncertainty channels, and cross-validation ensembles to achieve high classification and ranking performance. A comprehensive vetting catalog is produced, labeling thousands of TCEs and identifying planet candidates and CTOIs, thereby prioritizing targets for follow-up and maximizing planet yield. The work also rigorously analyzes data-quality limitations in TESS data, compares against existing classifiers, and presents ablation studies to validate the contribution of each new branch, offering a path toward scalable automated validation of exoplanets.

Abstract

We present ExoMiner++, an enhanced deep learning model that builds on the success of ExoMiner to improve transit signal classification in 2-minute TESS data. ExoMiner++ incorporates additional diagnostic inputs, including periodogram, flux trend, difference image, unfolded flux, and spacecraft attitude control data, all of which are crucial for effectively distinguishing transit signals from more challenging sources of false positives. To further enhance performance, we leverage multi-source training by combining high-quality labeled data from the Kepler space telescope with TESS data. This approach mitigates the impact of TESS's noisier and more ambiguous labels. ExoMiner++ achieves high accuracy across various classification and ranking metrics, significantly narrowing the search space for follow-up investigations to confirm new planets. To serve the exoplanet community, we introduce new TESS catalog containing ExoMiner++ classifications and confidence scores for each transit signal. Among the 147,568 unlabeled TCEs, ExoMiner++ identifies 7,330 as planet candidates, with the remainder classified as false positives. These 7,330 planet candidates correspond to 1,868 existing TESS Objects of Interest (TOIs), 69 Community TESS Objects of Interest (CTOIs), and 50 newly introduced CTOIs. 1,797 out of the 2,506 TOIs previously labeled as planet candidates in ExoFOP are classified as planet candidates by ExoMiner++. This reduction in plausible candidates combined with the excellent ranking quality of ExoMiner++ allows the follow-up efforts to be focused on the most likely candidates, increasing the overall planet yield.

ExoMiner++: Enhanced Transit Classification and a New Vetting Catalog for 2-Minute TESS Data

TL;DR

ExoMiner++ advances transit signal classification for TESS 2-minute data by incorporating new diagnostic inputs (e.g., periodogram, flux trend, difference image, momentum dumps) and leveraging multi-source learning with Kepler data. The model architecture integrates multiple input branches and uses Savitzky-Golay detrending, uncertainty channels, and cross-validation ensembles to achieve high classification and ranking performance. A comprehensive vetting catalog is produced, labeling thousands of TCEs and identifying planet candidates and CTOIs, thereby prioritizing targets for follow-up and maximizing planet yield. The work also rigorously analyzes data-quality limitations in TESS data, compares against existing classifiers, and presents ablation studies to validate the contribution of each new branch, offering a path toward scalable automated validation of exoplanets.

Abstract

We present ExoMiner++, an enhanced deep learning model that builds on the success of ExoMiner to improve transit signal classification in 2-minute TESS data. ExoMiner++ incorporates additional diagnostic inputs, including periodogram, flux trend, difference image, unfolded flux, and spacecraft attitude control data, all of which are crucial for effectively distinguishing transit signals from more challenging sources of false positives. To further enhance performance, we leverage multi-source training by combining high-quality labeled data from the Kepler space telescope with TESS data. This approach mitigates the impact of TESS's noisier and more ambiguous labels. ExoMiner++ achieves high accuracy across various classification and ranking metrics, significantly narrowing the search space for follow-up investigations to confirm new planets. To serve the exoplanet community, we introduce new TESS catalog containing ExoMiner++ classifications and confidence scores for each transit signal. Among the 147,568 unlabeled TCEs, ExoMiner++ identifies 7,330 as planet candidates, with the remainder classified as false positives. These 7,330 planet candidates correspond to 1,868 existing TESS Objects of Interest (TOIs), 69 Community TESS Objects of Interest (CTOIs), and 50 newly introduced CTOIs. 1,797 out of the 2,506 TOIs previously labeled as planet candidates in ExoFOP are classified as planet candidates by ExoMiner++. This reduction in plausible candidates combined with the excellent ranking quality of ExoMiner++ allows the follow-up efforts to be focused on the most likely candidates, increasing the overall planet yield.

Paper Structure

This paper contains 24 sections, 11 figures, 18 tables.

Table of Contents

  1. Introduction
  2. Machine Learning Classification of Transit Signals
  3. Existing Classifiers
  4. ExoMiner++
  5. Deep Learning Implementation
  6. Data Processing
  7. Data Preparation
  8. Preprocessing Pipeline
  9. Hyperparameter Optimization and Model Training
  10. Performance Study
  11. Evaluation Metrics
  12. Multi-Source Learning
  13. Multi-Sector Aggregation of the results
  14. Classification performance
  15. Ranking Performance
  16. ...and 9 more sections

Figures (11)

  • Figure 1: Classification Architecture.
  • Figure 2: Deep Learning Implementation of the Classification Design in Figure \ref{['fig:architecture']}. The dotted red boxes indicate the difference between ExoMiner++ and original ExoMinerValizadegan_2022_ExoMiner.
  • Figure 11: Score distribution versus orbital period (top) and planet radius (bottom) of non-exoplanet categories for the TESS+Kepler-Aggregate model.
  • Figure 12: PR-Curve and Confidence of different models.
  • Figure 13: Kepler versus TESS histograms of MES with a focus on values $>10$ for exoplanets (KP+CP) and AFPs (EB+FP+BD).
  • ...and 6 more figures