A Recursive Subdivision Technique for Sampling Multi-class Scatterplots
Results Parameter analysis for our method on the Person Activity data set. (a,b,c) Grid size influences the number of point samples. From left to right, the results have 5969, 2273, and 1217 points, respectively. (d,e,f) For a large λ, many outliers become visible, but overdraw happens in dense areas, while a small λ reduces overdraw but miss a few outliers. (g,e,h) A large τ shows too many outliers and regions of medium density are suppressed, while a small τ is more balanced but outliers are reduced. (i) When λ and τ both are large, the overdraw issue becomes severe while showing many outliers.
01-01-2020
