Abstract (eng)
This doctoral thesis aims to take a different look at stellar structure censuses in the Milky Way. Specifically, it aims to provide interpretable analysis methods to uncover both previously unknown stellar structures and new members of known stellar populations, providing astronomers with a more complete picture of the different stellar structures in the local Milky Way. The thesis contributions to the field are twofold: first, it introduces Uncover, an extended membership analysis technique that integrates known members of star clusters to search for yet undetected cluster members. Uncover is successfully applied to two different use cases, the recently discovered Meingast 1 stream, a Pleiades-age structure covering about 120 degrees of the sky, and the well-studied star-forming region rho Ophiuchus. For these two very different stellar structures, Uncover increased the number of members by tenfold and by about 200, respectively. Second, the thesis introduces Significance Mode Analysis (SigMA), an innovative clustering algorithm that studies the topological properties of the density field in multidimensional phase space. The application of SigMA to Gaia EDR3 data of the closest young association to Earth, the Scorpio-Centaurus (Sco-Cen) association, finds, for the first time, 48 co-moving and coeval clusters in Sco-Cen, many of them previously unknown. These 48 clusters are independently validated using astrophysical knowledge unknown to SigMA. Both Uncover and SigMA are formulated in domain-specific language, use expressive hyper-parameters, and allow for result validation to provide confidence in the results. With these tools, we seek to contribute to changing the current culture of blind acceptance of machine learning results and help astronomers build and modify models based on their expertise.