Mr. Biprateep Dey

Accurate photometric redshifts will be the key towards achieving the full scientific potential of upcoming large imaging only sky surveys like Rubin Observatory's LSST. We have developed deep learning-based photometric redshift estimation algorithms which provide state-of-the-art performance along with being significantly more interpretable and efficient than other deep learning-based methods. We have also developed methods to re-calibrate photometric redshift PDFs so that they are calibrated locally as a function of feature space. This ensures that the PDFs are well calibrated individually and not just as an ensemble. In this talk, we will discuss these new methods and how they may be used to study the physics of galaxies, localising transients, and much more.