Machine Learning

Joint image estimation and uncertainty quantification with R2D2 on Cygnus A VLA data

Towards a robust R2D2 paradigm for radio-interferometric imaging: revisiting DNN training and architecture

This paper adds the functionality of modelling epistemic uncertainty, introduces the novel DNN architecture, U-WDSR, and increases the robustness in R2D2 image reconstruction.

R2D2 image reconstruction with model uncertainty quantification in radio astronomy

This paper adds the functionality of modelling epistemic uncertainty in R2D2 image reconstruction.

Scalable Non-Cartesian Magnetic Resonance Imaging with R2D2

This paper brings the development of R2D2 to the field of medical imaging, specifically for magnetic resonance imaging with non-Cartesian sampling.

The R2D2 deep neural network series paradigm for fast precision imaging in radio astronomy

This paper proposes a new deep learning-based paradigm, R2D2, for fast precision imaging in radio astronomy, along with its variants R2D2Net and R3D3. The proposed paradigm is demonstrated on VLA simulated data, with comparison to CLEAN and state-of-the-art RI imaging algorithms, uSARA and AIRI.

CLEANing Cygnus A deep and fast with R2D2

This paper proposes a new deep learning-based paradigm, R2D2, for fast precision imaging in radio astronomy, along with its variants R2D2Net and R3D3. The proposed paradigm is demonstrated on real data of the radio galaxy Cygnus A, with comparison to CLEAN and state-of-the-art RI imaging algorithms, uSARA and AIRI.