Estimating Redshift from Galaxy Color Indices: Ridge Regression to Deep Neural Networks

This paper explores the use of machine learning models for estimating galaxy redshift fromphotometric color features derived from Sloan Digital Sky Survey (SDSS) data. We implemented andcompared three models: a simple Ridge Regression, a shallow neural network using basic color indices,and a deeper neural network with an expanded set of color combinations. While the Ridge model capturedsome linear trends, both neural networks performed noticeably better, with the deeper model achievingslightly higher accuracy. Although the models are relatively simple and trained on basic color inputs, theresults suggest that redshift prediction using photometric data is feasible without complex preprocessing ormassive datasets.