DR vs. HD-SDR for real-world data

Comparison of DR and HD-SDR using four different real-world data. DR methods are ordered from left to right based on how well clusters are separated. The samples are colored by their ground-truth labels for visual examination purposes. Note that the sharpening method significantly enhances the visual separability of clusters for the first three data sets, except for SUMAP as shown in the eighth column. Additionally, St-SNE and SUMAP exhibit more oversegmentation of clusters than SRP and SLMDS.