Speaker
Kwan Chuen Chan
(Sun-Yat Sen University)
Description
Accuracy photometric redshift (photo-z) estimation is crucial in imaging surveys. We present the photo-z estimation by the normalizing flow, a powerful deep learning method that can approximate complex probability distribution. We demonstrate that the method is able to give reliable photo-z estimation across a number of datasets. Besides accurate photo-z estimation, the characterization of the true redshift (true-z) distribution of a photo-z sample is also critical for unbiased cosmological parameter inference. By combining an improved self-calibration algorithm with the clustering-z method, we show that we can increase the true-z estimation accuracy, and extend the clustering-based method to higher redshift.
Primary author
Kwan Chuen Chan
(Sun-Yat Sen University)
