AAS 237 (402.01) - Mining TESS Data with Machine Learning

Transcript

1. Flare Statistics for Young Stars from a Convolutional Neural Network Analysis of TESS Data !1 Adina Feinstein NSF Graduate Research Fellow University of Chicago Benjamin Montet (UNSW), Megan Ansdell (NASA HQ), Brian Nord (UChicago/KICP/Fermi), Jacob Bean (UChicago), Maximillian Günther (MIT), Michael Gully-Santiago (UT Austin), Joshua Schlieder (NASA GSFC) @afeinstein20 March 306, 2020 ArXiv: 2005.07710 JOSS: 02347
2. The first million years of a star’s life dictates the conditions of its planets. !2
3. Previous studies used outlier metrics, which biases results towards identifying only the most energetic flares. (Chang, Byun, & Hartman, 2015) !3
4. Previous studies used outlier metrics, which biases results towards identifying only the most energetic flares. (Chang, Byun, & Hartman, 2015) !3
5. Previous studies used outlier metrics, which biases results towards identifying only the most energetic flares. (Chang, Byun, & Hartman, 2015) :( !3
6. Young stars light curves show significant modulation. Octans Columba !4
7. Young stars light curves show significant modulation. Octans Columba !4
8. Neural networks can be used when searching for signals with a characteristic shape. (Pearson et al. 2017) !5
9. Neural networks can be used when searching for signals with a characteristic shape. (Pearson et al. 2017) !5
10. !6 We compiled a sample of 3200 high probability young stars observed at 2-min cadence in Sectors 1-19. (Feinstein et al. 2020b)
11. !7 We used the TESS flare catalog from Günther et al. (2020) as our training, validation, and test sets. See Max Günther’s 
 iPoster 134.08 on flares 
 in TESS Years 1 & 2! (Feinstein et al. 2020a,b) CNNs are available @ MAST
12. The CNN is used as a sliding-box detector, where the probability will increase as the flare nears the center of the box. !8 https://github.com/afeinstein20/stella/
13. The CNN is used as a sliding-box detector, where the probability will increase as the flare nears the center of the box. !8 https://github.com/afeinstein20/stella/
14. We find a clear difference in flare rate at Gaia Bp -Rp = 2 (Teff ~ 4000K). !9 (Feinstein et al. 2020b)
15. Flares amplitudes and rates are higher the coolest stars across all ages, with noticeable evolution in other temperature bins. !10 (Feinstein et al. 2020b)
16. Across 1500 stars, there seems to be no preference for where flares happen. They happen everywhere! !11 Rotational Phase (Feinstein et al. 2020b)
17. Across 1500 stars, there seems to be no preference for where flares happen. They happen everywhere! !11 Rotational Phase (Feinstein et al. 2020b)
18. Across 1500 stars, there seems to be no preference for where flares happen. They happen everywhere! !11 Rotational Phase (Feinstein et al. 2020b)
19. Takeaways !12 • We have developed the first CNN for automated flare detection and does not require major light curve preprocessing.
 • M dwarfs have consistently high flare rates across the first 800 Myr, while hotter stars have fewer flares over time.
 at • M dwarfs with Teff < 3200 K have the highest normalized energy flares in the sample.
 • There is no phase dependence between flares and spot modulation, indicating consistent spot coverage on both hemispheres of these young stars. @afeinstein20 ArXiv: 2005.07710 JOSS: 02347