Aimee Schechter

Hi! I'm Aimee, an astrophysics postdoc at The University of Minnesota.

I am interested in the role galaxy mergers play in galaxy evolution through cosmic time.

I am broadly interested in galaxy evolution, particularly around cosmic noon. I have focused on studying galaxy mergers throughout my PhD with cosmological simulations and machine learning. Additionally, I have begun designing citizen science projects with the Zooniverse.

Here are some of the key projects I've been working on during my PhD; click the title to see the current version of the papers (I am in the process of making a few updates and incorporating feedback, but these will be on arxiv soon!):

• Beyond the Brightest: A Deep Learning Approach to Identifying Major and Minor Galaxy Mergers in CANDELS at z~1 I use CNNs to identify galaxy mergers. I am particularly interested in how we can use CNNs to find mergers beyond obvious, major mergers of spiral galaxies, especially at lower stellar masses. This is key to understanding the role of mergers at cosmic noon, as many galaxies don't yet look like galaxies in our local universe. I also want to build trust in these models by relating their outputs to astrophysical properties. Please see my submitted paper linked in the title, and stay tuned for an updated version in November 2025.
• Toward Complete Merger Identification at High Redshifts with Deep Learning This accepted paper for NeurIPS 2025 highlights my work on using interpretive techniques for a better understanding of neural networks for merger classification. Updated version with reviewer comments incorporated coming November 2025.
• Connecting Simulated and Observed Datasets with Domain Adaptation for Reliable Galaxy Merger Studies with CANDELS I am interested in building trustworthy CNNs, so that they can be applied to upcoming, unexplored datasets from Rubin and Roman reliably. To this end, I am working on using domain adaptation techniques such as Maximum Mean Discrepancy to go between simulated mock images and observations from HST CANDELS without a drop in accuracy. I am curious if domain adaptation could even help save computational time usually spent on radiative transfer in mock images. Once I trust the domain adaptation with well-understood CANDELS data, it's time to apply to new telescopes! I am aiming to accomplish this in my postdoc.
• Enhanced Star Formation and Black Hole Accretion Rates in Galaxy Mergers in IllustrisTNG50: I collaborated with astronomers at the CCA to investigate how mergers behave differently from nonmergers in IllustrisTNG50. We found mergers tended to have both excess sSFR and sBHAR.