From Simulations to Surveys: Domain Adaptation for Galaxy Observations

TL;DR

This paper introduces a domain adaptation pipeline that improves the transfer of galaxy morphology classification models trained on simulated data to real survey data by using advanced loss functions and top-k matching techniques.

Contribution

It presents a novel domain adaptation approach combining feature-level domain loss with top-k soft matching to enhance model transfer from simulations to real galaxy observations.

Findings

Target accuracy improved from ~46% to ~87% with adaptation.

Domain AUC near 0.5 indicates effective latent-space mixing.

Combines multiple loss functions for better domain alignment.

Abstract

Large photometric surveys will image billions of galaxies, but we currently lack quick, reliable automated ways to infer their physical properties like morphology, stellar mass, and star formation rates. Simulations provide galaxy images with ground-truth physical labels, but domain shifts in PSF, noise, backgrounds, selection, and label priors degrade transfer to real surveys. We present a preliminary domain adaptation pipeline that trains on simulated TNG50 galaxies and evaluates on real SDSS galaxies with morphology labels (elliptical/spiral/irregular). We train three backbones (CNN, $E(2)$-steerable CNN, ResNet-18) with focal loss and effective-number class weighting, and a feature-level domain loss $L_D$ built from GeomLoss (entropic Sinkhorn OT, energy distance, Gaussian MMD, and related metrics). We show that a combination of these losses with an OT-based "top_$k$ soft matching" loss that focuses $L_D$ on the worst-matched source-target pairs can further enhance domain alignment. With Euclidean distance, scheduled alignment weights, and top-$k$ matching, target accuracy (macro F1) rises from $\sim$46% ($\sim$30%) at no adaptation to $\sim$87% ($\sim$62.6%), with a domain AUC near 0.5, indicating strong latent-space mixing.