Space Science & Astrobiology Division Seminar Series
N245 Conference Room 215
Thursday, October 17, 2019 – 3:00 PM
Deep Learning for Heliophysics
Mark Cheung & Brad Neuberg
Lockheed-Martin
Abstract: NASA's Heliophysics division operates a fleet of spacecraft, the so-called Heliophysics System Observatory, to monitor the Sun's activity and how its changes drive space weather in interplanetary space and in the near-Earth environment. We'll present case studies of how a number of challenging problems encountered in Heliophysics can be tackled using deep learning. These include: (1) mega-Kelvin thermometry of the Sun's corona by using a deep neural network to solve a compressed sensing problem, (2) auto-calibration of solar EUV telescopes, and (3) virtual instruments to enhance the scientific return of scientific payloads. The work in this presentation was made possible by NASA's Frontier Development Lab, a public-private partnership between the agency and industry partners (including the SETI Institute, NVIDIA, IBM, Intel, kx & Lockheed Martin), whose mission is to apply artificial intelligence and machine learning to accelerate space science and exploration.
Mark Cheung bio:
Dr. Mark Cheung is a Staff Physicist at Lockheed Martin Solar & Astrophysics Lab and a Visiting Scholar at Stanford University. As Principal Investigator (PI) for the Atmospheric Imaging Assembly (AIA; http://aia.lmsal.com) onboard NASA’s Solar Dynamics Observatory mission, PI for NASA's Heliophysics Grand Challenges Research grant, and mentor for NASA’s Frontier Development Lab, he leads teams of scientists and engineers who operate space telescopes, perform data mining and data analysis on terabyte- and petabyte-scale data archives, develop massively parallel numerical simulation codes, and apply machine learning techniques for scientific discovery and space exploration. In 2017 he was the recipient of the Karen Harvey Prize awarded by the American Astronomical Society. Committed to STE(A)M education and outreach, he has appeared in television documentaries and given public lectures in venues such as Griffith Observatory in Los Angeles.
Brad Neuberg bio:
Brad Neuberg is a a full stack machine learning engineer, with 15+ years experience across startups, open source, and companies like Google and Dropbox. At NASA's Frontier Development Lab (FDL), he applies neural nets to such problems as satellite auto-calibration; image encoder/decoders for virtual telescopes; using GANs for scientifically useful embeddings; etc. Before NASA FDL, he was on the Machine Learning team at Dropbox, doing R&D to ship ML-powered products to millions of users. Recent work at Dropbox included launching features that automatically run advanced deep learning models on billions of images; creating an industry leading OCR pipeline with deep learning; and doing R&D for ML-powered features like meeting assistants and intelligent file systems.