[Physics-talks] PhD Thesis Defense: Sogol Sanjaripour 10/19/2026

[Melia Timo]

*Learning Representations of Galaxies: **From Self-Organizing Maps to
Foundation Models*

Sogol Sanjaripour
PhD Thesis Defense
*Monday, October 19 · 11:00 AM (Pacific Time)*
*In Person:* MRB Conference Room, First Floor
*Online:* Join via Zoom
<https://ucr.zoom.us/j/91455184475?pwd=ElhoRNckcr8olvYmaNIbAcuDfBQk5O.1>

Upcoming surveys such as Euclid and Roman are transforming our view of the
universe by providing imaging and photometric observations of enormous
galaxy populations. However, spectroscopic measurements needed to constrain
many of their physical properties remain limited. This raises an important
question: how much information about galaxy evolution and physical
properties is encoded in astronomical images, and how effectively can
machine learning extract it? In my dissertation, I investigate this
question through unsupervised and self-supervised representation learning,
from Self-Organizing Maps to transformer-based foundation models, with
applications to galaxy populations, scaling relations, and active galactic
nuclei in dwarf galaxies.

For those joining in person, feel free to arrive 10–15 minutes early to
enjoy some snacks before we begin :)

Everyone is welcome to join in person or online! Hope to see you there!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ucr.edu/pipermail/physics-talks/attachments/20261016/cb863716/attachment.htm>