Exploring the universe with AI Kevin Schawinski Institute for - - PowerPoint PPT Presentation

@kevinschawinski

ETH black hole group

Grüp Bœgg Negar Politecnic da Zürig

Exploring the universe with AI

Kevin Schawinski

Institute for Particle Physics and Astrophysics ETH Zurich

how can machine learning/ artificial intelligence help us understand the universe?

“BIG DATA”

GalaxyGAN: de-noising and feature reconstruction PSFGAN: point source subtraction Generative models: data-driven exploration

Original Image Degraded Image Artificial Degrading

Data Prep. Training of GAN

Recovered Image Generator Original Image Discriminator (Original Image, Degraded Image) or (Recovered Image, Degraded Image)

generative adversarial network for overcoming limitations in astrophysical images

Schawinski+17

• riginal

degraded GAN recovered deconvolved PSF=2.5”, 10σ

Training Architecture Discriminator Generator Preprocessing Recovered Original Original + AGN

PSFGAN, Stark+ submitted

Dominik Stark

PSFGAN, Stark+ submitted

PSFGAN, Stark+ submitted

Less sensitive to PSF changes Better at recovering features

• riginal data

encoder latent space z decoder reconstructed data

Dennis Turp

z’ = a×z1 + b×z2 z1 z2

z’ = a×z1 + b×z2 z1 z2

• riginal data

encoder latent space z decoder reconstructed data

• riginal galaxy

reconstructed galaxies with SSFR changed in latent space

z SSFR z

age

• riginal face

reconstructed faces with age changed in latent space

changing SSFR in latent space changing bulge-to-disk in latent space

machine learning can help us do better science by better understanding the data we have, and will get in the future

go to space.ml to try out our projects!