2026-11-04 –, Seminar Room 2
As in many other disciplines, also in astronomy Machine Learning (ML) is becoming more and more important. Current workflows often force researchers to repeatedly extract and preprocess data in specific formats for using them in their individual ML tools. We present a concept that extends the International Virtual Observatory Alliance (IVOA) vision of interoperability to ML models. Our goal is to enable an easy integration of ML models into existing astronomical tools. By using standardized model representations such as ONNX, our approach enables models to be tested on diverse datasets without significant additional effort. This standardization makes models interchangeable, allows fair comparisons, accelerates iterative development, and simplifies the extension of tools with individual functionalities. Finally, we demonstrate a prototype that integrates various custom-built ML models into the TOPCAT software, showcasing practical applications such as automatic classification or redshift estimation.
Master's student in the Astroinformatics group at HITS gGmbH.