Table 2.
MDN performance evaluation, without any clipping for the average and rms, without any threshold on branch membership probabilities.
IGMM | Photometry | ⟨Δz⟩ | rms(Δz) | 3σ outliers | ⟨Δz⟩, | rms(Δz), | 3σ outliers, | rms(Δz), | rms(Δz), |
---|---|---|---|---|---|---|---|---|---|
implementation | (all) | (all) | (all) | range1 (a) | range1 (a) | range1 (a) | range2 (b) | range3 (c) | |
Fully unsup. | griz, W1, W2 | 0.0152 | 0.2174 | 3.08% | 0.0007 | 0.0177 | 0.28% | 0.0988 | 0.0945 |
Spec. class | griz, W1, W2 | 0.0111 | 0.2069 | 1.31% | 0.0006 | 0.0167 | 0.41% | 0.0822 | 0.0783 |
Spec. class (d) | griz, W1, W2 | 0.0356 | 0.2300 | 1.35% | 0.0110 | 0.0260 | 0.71% | 0.0953 | 0.0903 |
Redshift (zs) | griz, W1, W2 | 0.0176 | 0.2131 | 3.21% | −0.0009 | 0.0174 | 0.38% | 0.0896 | 0.0873 |
Spec. class, zs | griz, W1, W2 | 0.0047 | 0.1990 | 2.66% | 0.0036 | 0.0181 | 0.57% | 0.0675 | 0.0664 |
Spec. class | ugriz, W1, W2 | 0.0135 | 0.1592 | 1.62% | 0.0007 | 0.0160 | 0.23% | 0.0601 | 0.0611 |
Notes. Spectroscopic sample for all IGMM implementations containing stars, galaxies and quasars.
Fully unsup.: Fully unsupervised IGMM implementation. Spec. class: uses spectroscopically classified objects (e.g., stars, galaxies, and quasars) for those which SDSS provides spectroscopic information (i.e., ≈2% of the photometric data set) to classify the data. Redshift (zs): uses the spectroscopic redshift of objects from ≈2% of the photometric data set to classify the data. Spec. class, (zs): uses both, spectroscopically classified objects and the spectrsocopic redshift of objects from ≈2% of the photometric data set to classify the 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.