Siou-Yu Chang,(Department of Physics, National Chung Hsing University, 40227, Taichung, Taiwan) Yu-Yen Chang,(Department of Physics, National Chung Hsing University, 40227, Taichung, Taiwan)
We use deep learning techniques to analyze galaxy parameters with Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) Public Data Release 2 (PDR2) data. With python packages (e.g., Keras sequential model), we predict y-band data from g-,r-,i-, and z- band photometry, as well as estimate redshifts of galaxies from g-, r-, i-, z-, and y- band photometry. We optimize our inputs by normalizing fluxes, magnitudes, and colors. As a result, flux with individual galaxy normalization can obtain the best performance of R squared values (>0.95 for photometry estimation and >0.70 for redshift estimation). We separate the sample to different redshift bins as various classifications, but do not find better results than our regression models. In the future, we will also include five-band HSC imaging as input data, and apply our methods to global galaxy properties.