|
|
|
Figure 10
The completion and spectral errors for DS2 associated with each of the discussed algorithms: random raster sampling, data-driven raster sampling (RISS), and data-driven CUR raster sampling (CURISS). Solid lines show the mean across 100 trials for CURISS and 10 trials for random raster sampling and RISS. Shaded bands represent ±1 standard deviation in the log-scale. The standard deviations show that the variance is low and the probability that RISS or CURISS fails is very small. |

journal menu![[Figure 10]](bon5003fig10.jpg)
access


