Poster Abstract

P4.4 Youngjoo Kim (Bernoulli Institute for Mathematics, Computer Science, and Artificial Intelligence, RUG)

Theme: Data visualisation from line plots to augmented & virtual reality

Analysis of Stellar Chemical Abundance Space with F-LMDS

Dimensionality reduction (DR) is an effective method to visually analyze astronomical surveys with large number of dimensions and observations. A previous study using the state-of-the-art DR technique, t-distributed stochastic neighbor embedding (t-SNE) shows interesting substructures in the 2D abundance-space using a number of stellar abundances as input. However, a more recent study using Filtered Landmark Multidimensional Scaling (F-LMDS) is used as an alternative to t-SNE. This method shows potentially more interesting substructures in the 2D abundance-space than t-SNE when analyzing the second data release of GALactic Archaeology with HERMES survey (GALAH DR2) cross-matched with Gaia DR2. Here we aim to explain the significance of the results for the first time by color-coding and labelling the substructures and describing their connections to alpha-elements and other properties, such as the 6D phase-space coordinates and velocity information, surface gravity, and temperature. These results show that F-LMDS is an effective DR method for chemical-tagging studies.