Example of running PCA on the DECam data set for GW170817 KN. Top: we give the observation time, visit/exposure number, and CCD number. Then, we show the effective exposure time, and “EVR” means explained variance ratios of the first and second principal components. The title also lists the PSF flux S/N on the difference image. Left: the first (PC1) and second (PC2) principal component values of sources in the transformed space. Those points are all remaining sources after flag filtering (Section 4.1). The target (KN) is marked by an orange point. Middle: the coefficient of each feature along each principal component direction (i.e., the projection on the principal component vector). Hollow markers indicate the negative values (sign flipped). The reference horizontal dotted line denotes one divided by the square root of the number of features (﹩\sqrt{1/15}\sim 0.26﹩), showing the effectiveness of each feature. Right: the feature quantiles of the target (i.e., the fraction of sources that have a value for this feature smaller than or equal to that of the target). The horizontal dashed line gives the PC1 quantile of the target among sources, showing the effectiveness of candidate selection (R/B).