Open clusters in the machine learning era

ABSTRACT:

Open Clusters are gravitationally bound associations of stars that have been born together, thus sharing the same age and chemical composition. For these reasons, they represent the ideal laboratories to perform studies in a broad range of fields such as Galaxy structure and its evolution, stellar evolution, and gravity. The advent of the Gaia mission, culminating in Gaia DR3 with over 1.8 billion sources, has significantly expanded our ability to detect and characterize open clusters, resulting in more than 10,000 new candidates by 2023.

As astronomy transitions into the Big Data era, the integration of Machine Learning algorithms becomes crucial for the automated detection and analysis of these clusters. In this seminar, I will explore existing tools, highlighting their advantages, limitations, potential pitfalls, and their implications on research outcomes.

I will also discuss some methods useful to perform sanity checks on these algorithms (e.g. using t-Distributed Stochastic Neighbour Embedding; t-SNE maps). Furthermore, I will present an innovative approach using an Artificial Neural Network (Machine Learning algorithms inspired by biological neural networks) in conjunction with Quad-Trees (tree data structure mapping 2D data to 1D) to determine the age, extinction, and distance of open clusters.