Peer-Reviewed Scientific Article / Machine Learning And Historical Astronomical Plates
Use of machine learning to enhance detection of transient astronomical phenomena in historical observatory images
Peer-reviewed Scientific Reports study training a supervised classifier on 250 expert-labeled transient image pairs and applying it to 107,875 previously identified candidates. The authors report that nuclear-window and Earth-shadow patterns persist after model-based artifact control, while the physical nature of the candidates remains unresolved.