QuantifAI: Scalable Bayesian uncertainty quantification with data-driven (learned) priors

Github

QuantifAI is a PyTorch-based open-source radio interferometric imaging reconstruction package with scalable Bayesian uncertainty quantification relying on data-driven (learned) priors.

The methods developed and implemented in QuantifAI are also being ported to the C++ PURIFY and SOPT codes for exascale computing.

code

Publications

Scalable Bayesian uncertainty quantification with data-driven priors for radio interferometric imaging

Preprint PDF Code DOI

Talks

Towards learned exascale computational imaging (LEXCI) for radio astronomy

Mar 2025 University of Cambridge
PDF Code Code Code

Learned Exascale Computational Imaging (LEXCI)

Oct 2024 Leicester
PDF Code Code Code

Towards learned exascale computational imaging for the SKA

Sep 2024 Virtual
PDF Code Code Code

Scientific AI in cosmology

May 2024 Chania
PDF Code Code Code Code Code Code Code

Learned Exascale Computational Imaging (LEXCI): Update

Oct 2023 Bristol
PDF Code Code Code

Scientific machine learning in astrophysics: machine learning for physics; physics for machine learning

Sep 2023 Harwell
PDF Code Code Code

Learned Exascale Computational Imaging (LEXCI)

May 2023 University of Leicester
PDF Code Code Code

Learned Exascale Computational Imaging (LEXCI) overview

Jul 2022 Met Office
PDF Code Code Code

Learned Exascale Computational Imaging (LEXCI) overview

Apr 2022 UCL
PDF Code Code Code