AstroAI Lunch Talks - April 14, 2025 - Sogol Sanjaripour
14 Apr 2025 - Joshua Wing
The video can be found here: https://www.youtube.com/watch?v=BNEvuJOdlNQ
Speaker: Sogol Sanjaripour (UC Riverside)
Title: Manifold Learning: Selection of Different Galaxy Populations and Scaling Relation Analysis
Abstract: The increasing volume of data from large astronomical surveys necessitates advanced analysis techniques to effectively manage high-dimensional datasets. In this talk, I will discuss the application of unsupervised machine learning, specifically the Self-Organizing Map (SOM), to study the spectral energy distribution (SED) space of galaxies. Our goal is to bridge photometry and spectroscopy by training an ML model on photometric data to predict spectroscopic features, reducing the high cost and time demands of spectroscopy.
Using data from the CANDELS and KECK MOSDEF surveys at z ~ 1.5 and z ~ 2.2, we trained a SOM on photometric data and mapped spectroscopic data onto it to analyze sample selection biases and galaxy properties. Our findings reveal that MOSDEF AGNs preferentially occupy the more massive regions of the SOM, confirming known selection biases toward high-mass, less dusty galaxies. We also examined metallicity variations across the SOM, showing that more massive galaxies exhibit lower [O III]/Hβ and [O III]/[O II] ratios and higher Hα/Hβ ratios, consistent with the mass-metallicity relation. These results highlight the potential of SOMs in uncovering trends in multidimensional datasets and improving AGN and galaxy population studies.