Searching for Habitable Worlds Using Artificial Intelligence

Title: Searching for Habitable Worlds Using Artificial Intelligence

Speaker: Prof. Jian Ge (University of Florida)
Time 10:30 am, Dec. 13 (Wednesday)  
Location: Lecture Hall, 3rd floor 
Abstract: Detecting habitable worlds around nearby stars and searching for life among them is one of the main goals of science. However, it is extremely challenging to detect them with the ground-based Doppler spectroscopy and space-based transit photometry as their signals are extremely weak. We launched the Dharma Planet Survey (DPS) using the TOU very high precision Doppler spectrograph at the fully dedicated 50-inch Automatic Dharma Endowment Foundation Telescope (DEFT) on Mt. Lemmon in 2016 to search for habitable planets around nearby FGKM dwarfs. To date, more than 80 survey targets have been observed with more than 10 times each. Seven low-mass planet candidates have been detected. In the meantime, we are developing deep neural networks to capture weak long-period transit signals from habitable Earth-like planets in the Kepler photometry data, taking advantage of the newly established NSF deep learning (DL) center at UF. An early DL model was able to detect over 20,000 weak metal absorption lines among SDSS 50,000 quasar spectra at better than 90% accuracy in an extremely short time of only 20 seconds. This DL method is being implemented in Kepler data and will be implemented in DPS spectroscopy data to significantly boost search speed to detect weak habitable planet signals and optimize DPS operation efficiency. Early results will be presented.  

 

  

  


附件下载: