Speaker
Description
One of the crucial keys in the cosmological studies is the estimation of an accuracy redshift of a large number of galaxies. Sometimes, the spectroscopic sample used as training sample for ML approaches doesn't cover the same magnitude and color space as the target sample. This issue raises doubts about the confidence of the photometric redshift provided by the algorithms.
In this talk, we present the effect of using complete or incomplete spectroscopic training samples to determine the photo-zs by DNF algorithm. We compare the photo-zs estimate for the validation sample using both training samples. We provide a new method for determining the level of confidence in the photo-z values and the incompleteness assessment of the results. Finally, we compare the DNF photo-zs with templates methods.
| Main Topic | Classification and regression |
|---|---|
| Participation mode | In person |
