About me

EWASS 2020, AI in astrophysics session

Contact: david.cornu@observatoiredeparis.psl.eu
My LinkedIn profile (with education and experience details)

I am a French numerical astrophysicist, presently employed as an AI Fellow at PSL University (Paris Science Lettre University) in the context of the cross-disciplinary Data Science program.
I work at the Paris Observatory as a member of the LERMA (Laboratoire d’Etudes du Rayonnement et de la Matière en Astrophysique et Atmosphères).

My current research topic revolves around developing state-of-the-art Machine Learning (ML) approaches and tools for analyzing massive datasets from modern giant interferometers (LOFAR, ALMA, NenuFAR, MeerKAT, SKA, …) in the context of the MINERVA project.
I was noticeably the PI of the MINERVA team competing in the SKA Science Data Challenge 2, for which I developed a highly customized 3D-YOLO network. The MINERVA team won first place in the challenge.
Press release and articles about the victory: SKAO (& Contact 9 pp16-17), CNRS, OBSPM, OCA, GENCI, ActuIA, and others …

Overall, I am interested in the ML method’s capabilities to solve existing and upcoming astrophysical problems and how the specific properties of astronomical datasets challenge usual ML approaches.

Previous and present primary studies:
– Object detection, classification, and parameter extraction in radio-astronomical
datasets using modern CNN architectures
– Extinction mapping of the Milky Way using CNNs
– Young Stellar Objects (YSOs) classification from Infrared surveys, mainly using ANNs
– Various investments in creating/modifying ML tools for astronomical data/problems

Other side topics:
– Model inversion of numerical simulations of the Epoch of Reionization
– Visualization tools for astronomical data
– Neural accelerators for simulations acceleration or emulation

See my publications page.

For my studies, I developed a state-of-the-art Deep Learning framework called CIANNA (Convolutional Interactive Artificial Neural Networks by/for Astrophysicists), which is already competitive with widely adopted DL libraries (see Tools and Codes). CIANNA was my primary tool to work on the SKA SDC2.