Figure 5.
Coefficient histograms for SVD-CCC with missing data and noise. (a, b): 10% missing data, SNR 0.2 and 0.02, respectively. (c): 30% missing data, SNR 0.2. In (a), coefficient 2 yields perfect 3-class separation. Clustering using k-means gives only 2-class separation in (b) and fails to find any significant clusters in (c).