Automated Identification and Segmentation of Neutral Hydrogen (HI) Sources in the CRAFTS Survey Using Deep Learning
A deep learning-based method is introduced for accurately identifying and segmenting neutral hydrogen (HI) sources from 3D spectral data cubes of the CRAFTS survey, achieving high recall (91.6%) and accuracy (95.7%).
R2D2 Image Reconstruction with Model Uncertainty Quantification in Radio Astronomy
R2D2's reconstruction method utilizes deep learning for high-quality imaging in radio astronomy, with a focus on uncertainty quantification.
Novelty Detection Framework for Radio Astronomy Data Using Signatures
SigNova introduces a novel semi-supervised framework for detecting anomalies in streamed data, with a focus on radio-frequency interference (RFI) in radio astronomy. The approach leverages signature transforms and Mahalanobis distances to improve RFI detection.
The R2D2 Deep Neural Network Series Paradigm for Fast Precision Imaging in Radio Astronomy
The author introduces the novel deep learning approach of the R2D2 algorithm to address scalability challenges in radio astronomy imaging, combining elements of PnP algorithms and matching pursuit. The core thesis is that R2D2 offers high precision and fast imaging capabilities through a series of residual images generated by DNNs.
VisRec: Semi-Supervised Radio Interferometric Data Reconstruction
VisRec proposes a semi-supervised learning approach for radio interferometric data reconstruction, leveraging both labeled and unlabeled data effectively.