Detecting outliers and learning complex structures with large spectroscopic surveys - a case study with APOGEE stars

摘要

In this work we apply and expand on a recently introduced outlier detectionalgorithm that is based on an unsupervised random forest. We use the algorithmto calculate a similarity measure for stellar spectra from the Apache PointObservatory Galactic Evolution Experiment (APOGEE). We show that the similaritymeasure traces non-trivial physical properties and contains information aboutcomplex structures in the data. We use it for visualization and clustering ofthe dataset, and discuss its ability to find groups of highly similar objects,including spectroscopic twins. Using the similarity matrix to search thedataset for objects allows us to find objects that are impossible to find usingtheir best fitting model parameters. This includes extreme objects for whichthe models fail, and rare objects that are outside the scope of the model. Weuse the similarity measure to detect outliers in the dataset, and find a numberof previously unknown Be-type stars, spectroscopic binaries, carbon rich stars,young stars, and a few that we cannot interpret. Our work further demonstratesthe potential for scientific discovery when combining machine learning methodswith modern survey data.