Abstract

Poster - Plenary

Galaxy Classification Using Machine Learning with the SpecLess Model

Uzay Aydın, Mehmet Tanrıver
Erciyes University

The rapid growth of large-scale photometric surveys demands robust and scalable methods for galaxy classification that do not rely on spectroscopic information. In this work, we present SpecLess, a machine learning-based framework designed to classify galaxies using only photometric data. The model leverages multi-band magnitudes and colour indices to distinguish between major galaxy populations, including star-forming systems, active galactic nuclei (AGN), and different morphological types. SpecLess is built as a modular and hierarchical pipeline, combining binary and multi-class classification strategies to improve interpretability and performance. The framework is trained and validated on large samples derived from the Sloan Digital Sky Survey (SDSS), where reliable spectroscopic labels are available for benchmarking purposes. Our results demonstrate that the model achieves high classification accuracy, reaching up to ~98% in controlled test samples, while maintaining strong generalization capabilities across different galaxy populations. The absence of spectroscopic dependence makes SpecLess particularly suitable for next-generation surveys such as the Vera C. Rubin Observatory LSST, where the volume of photometric data will far exceed spectroscopic coverage. Ongoing work focuses on improving robustness, domain adaptation across surveys, and probabilistic outputs for population studies. SpecLess provides a scalable and survey-ready solution for automated galaxy classification in the era of data-intensive astrophysics.