Ashod Khederlarian

In this project, we trained a JAX-implemented neural network on DESI early data release to predict the strengths of emission lines of galaxies from their stellar continua.

Most of the light from galaxies comes from stars, which produce a smooth spectrum, and from gas, which emit a lot of light at specific wavelengths, called emission lines. This project shows that these two sources are strongly correlated. This is not surprising - emission lines are produced when the gas is ionized by the light emitted from the stars. Also, the starlight holds information on the history of star formation of the galaxy, which is what determines the content of the gas.

The top-left panel in the figure shows a 2D projection of the starlight (using UMAP), with 1,000 galaxies being represented as scatter points color-coded with the strength of the strongest emission line, H-alpha. There is a clear trend from weak to strong H-alpha emission. The galaxies with strong H-alpha emission are the ones with the youngest stars (which have blue continua), and the galaxies with weak H-alpha emission are the ones with the oldest stars (which have red continua). The top-right panel shows the spectrum of a particular galaxy, and the bottom four panels show some of its emission lines. This galaxy has a blue spectrum (the flux is strongest at shorter wavelengths), and it also has strong emission lines.

You can checkout an interactive version of this plot at this link (it might take some time to load). By hovering over the points in the scatter plot of the top-left panel, you will find that galaxies with a red spectrum (more flux at longer wavelengths) will have weak emission lines.

This project aims to improve distance (redshift) estimates of far-away galaxies (up to redshift ~ 2) by leveraging pixel-level information from space-based images. For nearby galaxies, it has been shown that the shape of the galaxy can improve distance estimates. This should also be true for galaxies that are further away, but high-resolution images from space telescopes are required.

To test this, we have used contrastive learning (a machine learning method that uses deep neural networks) on Hubble Space Telescope (HST) images in the CANDELS fields to compress them into a low-dimensional space. In this space, galaxies which have similar shapes (up to rotation) and colors will be close to each other. In the figure above, we show as an example 1 galaxy (left column) and its 5 nearest neighbors (columns 2-6) in the low-dimensional space. The rows correspond to different HST imaging filters (wavelengths). The neighbors have similar shapes to the original galaxy, and also have similar relative fluxes in the different filters.

After obtaining a low-dimensional representation, we trained a different neural network to predict distances from it. We saw minor improvements in performance, but we think we can do better by using transfer learning (with ground-based images) and a more sophisticated neural network architecture, so stay tuned!

I am Lebanese-Armenian born and raised in the beautiful country of Lebanon. I enjoy rock climbing, playing football (soccer), and photography. I love exploring new places.