This paper presents a comprehensive methodological framework for cross-survey transfer learning in radio astronomy artifact detection, addressing challenges like domain shift, label scarcity, and emerging artifact classes.
Key findings
Proposes a domain-adaptive architecture combining self-supervised pre-training, few-shot adaptation via meta-learning, and anomaly detection.
Addresses frequency-dependent artifact signatures, varying resolution and sensitivity across surveys, and emergence of novel artifact classes.
Outlines detailed experimental plans involving ASKAP EMU, MeerKAT GPS, and LOFAR LoTSS surveys.
Limitations & open questions
The proposed framework requires substantial computational resources for training and adaptation.
The effectiveness of the framework in real-world deployment may be limited by the availability of labeled data for new surveys.