Breaking the degeneracy in stellar spectral classification from single wide-band images

Open Access

Table 2

Classification metrics for the PCA+MLP, CNN+MLP, and SVM+PSF models.

Model	F1	Accuracy	Top-two accuracy
PCA+MLP	0.366	0.370	0.757
CNN+MLP	0.385	0.391	0.746

$SVM + {PSF}_{S_{1}}$ ${\rm{SVM}} + {\rm{PS}}{{\rm{F}}_{{{\cal S}_1}}}$	0.392	0.410	0.755
$SVM + {PSF}_{S_{4}}$ ${\rm{SVM}} + {\rm{PS}}{{\rm{F}}_{{{\cal S}_4}}}$	0.506	0.512	0.873
SVM+PSF_GT	0.546	0.549	0.910

Notes. $SVM + {PSF}_{S_{1}}$ ${\rm{SVM}} + {\rm{PS}}{{\rm{F}}_{{{\cal S}_1}}}$ stands for the SVM+PSF classifier that uses the similarity features computed with an approximate PSF model trained on the 𝒮₁ dataset, which has a relative error of 2.4%. Analogously for the 𝒮₄ dataset with 500 stars and a relative error of 1%. The SVM+PSF_GT row uses the ground truth PSF to compute the similarity features.

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.