September 2021
Dr. Hamed Valizadegan, (Ph.D.), Senior Scientist, Machine Learning (USRA) and Mr. Miguel Saragoca- Martinho, Associate Scientist, Data Science (USRA) were awarded a ROSES TESS Guest Investigator fund to investigate the application of their developed deep neural network (DNN) classifier to classify Transiting Exoplanet Survey Satellite (TESS) Cycle 4 Threshold Crossing Events (TCEs). A proposal titled, "Vetting And Ranking Tess Cycle 4 TCEs Using A Novel And Accurate Deep Neural Network", submitted in response to the NASA Research Announcement NNH20ZDA001N for participation in the TESS Cycle 4 Guest Investigator Program, has been awarded for funding. TESS Cycle 4 observations will produce extensive datasets that will result in thousands of additional candidate exoplanet transit signals from which we can expect hundreds of new planet candidates. The established method to vet these signals is based on a semi-manual vetting process that starts with a triage code and follows with a more accurate manual vetting. Instead, this proposal entails the use of an accurate, reliable, and explainable deep neural network (DNN) designed by mimicking how human vetters utilize all unique elements of a data validation report in order to identify different types of false positives before vetting a TCE. The new DNN, called ExoMiner, provides an accurate disposition score so that the domain scientists can focus on the most likely planet candidates for follow-up study.
The project is a collaboration between USRA, NASA Ames Research Center and the SETI Institute. Team collaborators include: Dr. Hamed Valizadegan (PI) and Miguel Martinho (Co-I) from USRA, Dr. Douglas Caldwell (Co-I), (Ph.D.), (SETI Institute), Dr. Joseph Twicken (Ph.D.), (SETI Institute), and Dr. Jeffrey Smith (Ph.D.), (SETI Institute), and Dr. Jon Jenkins (Ph.D.), (NASA Ames).