Photometric redshift-aided classification using ensemble learning

Home

All issues

Volume 666 (October 2022)

A&A, 666 (2022) A87

Full HTML

Open Access

Table 3

Regression metrics for photo-z predictions level one and two (top rows), regression metrics for the SDSS photo-z prediction and this work, only for galaxy sources (bottom rows).

Algorithm	Level one				Level two
	Bias	NMAD	R²	Outlier fraction	Bias	NMAD	R²	Outlier fraction
XGBoost	0.015	0.021	0.899	0.0376	0.013	0.018	0.909	0.0290
CatBoost	0.015	0.021	0.905	0.0347	0.013	0.018	0.914	0.0272
LightGBM	0.015	0.021	0.903	0.0344	0.013	0.018	0.915	0.0270
Ensemble	0.014	0.019	0.906	0.0342	0.0124	0.018	0.916	0.0265

SDSS (galaxies only)	_–	–	–	–	0.014	0.020	0.714	0.0226
This work (galaxies only)	–	–	–	–	0.013	0.019	0.912	0.0081

Notes. The metrics for XGBoost, CatBoost and LightGBM are representative of the testing data. Ensemble is the average of the predictions given by the base learners, with models built in the k-fold procedure using the training data.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.