Table 3
Performance of the TPZ algorithm, estimated by splitting our spectroscopic sample (see Sect. 3.2) into train and test files.
| Sample | Point like | Extended | TPZ parameters | ||||
| σnmad/η (%) | ⟨error⟩ | σnmad/η (%) | ⟨error⟩ | Nrandom | NTrees | Natt | |
|
|
|||||||
| SDSS | 0.08/27.0 | 0.33 | 0.06/18.0 | 0.21 | 6 | 8 | 7 |
| SDSS+WISE | 0.06/17.4 | 0.25 | 0.06/13.0 | 0.20 | 8 | 10 | 8 |
| SDSS+WISE+NIR | 0.05/13.7 | 0.23 | 0.04/9.0 | 0.18 | 6 | 8 | 12 |
| SDSS+NIR | 0.06/20.0 | 0.27 | 0.05/11.5 | 0.19 | 8 | 10 | 10 |
Notes. The accuracy of the photometric redshifts is quantified by estimating the normalized absolute median deviation, σnmad and the percentage of outliers, η. The median error of the photometric redshift for each subsample is shown. The values of the TPZ parameters we used for each subsample are also presented.
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.