Repairing errors of (from top to bottom) (1) PCA on spectra with 10 <S/N ≤ 20; (2) PCA on spectra with S/N> 20; (3) PCA on spectra with 0 <S/N ≤ 10; (4) RBM on spectra with 0 <S/N ≤ 10; (5) RBM on spectra with S/N> 20; and (6) RBM on spectra with 10 <S/N ≤ 20. The repairing error is the average difference between the repaired spectra and the true spectra in the missing region. From the figure we find that the repairing errors given by RBM are significantly smaller than those given by PCA. Furthermore, RBM performs worst on data set D3 (S/N ≤ 10), while PCA performs best on D3, possibly because PCA is a linear method, as opposed to RBM which is a non-linear method. Thus, PCA performs best on D3 whose principal feature is line.
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.