Researchers find thousands of hidden planet candidates in TESS data
A team led by a Princeton graduate student reprocessed TESS year-one data down to 16th magnitude and ran a machine learning pipeline on roughly 84 million brightness records across about 54 million stars. Using two random forest classifiers trained on confirmed planets, eclipsing binaries and injected signals, they identified 11,554 planet candidates, 10,091 of them previously unseen, and validated one hot Jupiter with the Magellan telescope. The list includes 411 single-transit events and candidates with periods from 12 hours to about 27 days, and the authors estimate 3,000 to 5,000 may be real planets worthy of follow-up.
Machine learning found 11,554 planet candidates in TESS year-one data.
Context
TESS collected its first year of data in 2018 and most faint-star records were not…
The full analysis
19 dimensions on this story — world impact, market read, and what happens next.
- Full ContextLocked
- Affected SectorsLocked
- Stock ImpactLocked
- Economic IndicatorArchive-data reanalysis with machine…Locked
- Investor RelevanceLocked
- Professional RelevanceLocked
- Watch PointsCount of candidate confirmations from…Locked
- Probability of ChangeLocked
- Debate PointsLocked
- Historical ParallelEarly forecast papers predicted hidden…Locked
- Prerequisite KnowledgeLocked
- Follow-up QuestionsHow many of the new candidates will be…Locked
- Pros & ConsAstronomers: gain many new targets…Locked