Sample points based on distance matrix

Sample points based on distance matrix so that distance between any pair is not less than a fixed radius. This will not preserve density, but does preserve local outliers.

condense(d, radius, seed = 1, ...)

Arguments

d

object of class 'dist'
radius

(numeric) radius around a point
seed

(positive integer) A seed. Default is 1.
...

Arguments for frNN

Value

Index of condensed points

Examples

library("magrittr")
dataset <- rbind(
  MASS::mvrnorm(1000, mu = c(0, 0), Sigma = matrix(c(1, 0.3, 0.3, 1), 2))
  , MASS::mvrnorm(100, mu = c(10, 10), Sigma = matrix(c(1, 0.3, 0.3, 1), 2))
  , MASS::mvrnorm(10, mu = c(5, 5), Sigma = matrix(c(1, 0.3, 0.3, 1), 2))
  , MASS::mvrnorm(10, mu = c(0, 8), Sigma = matrix(c(1, 0.3, 0.3, 1), 2))
  , MASS::mvrnorm(10, mu = c(7, 0), Sigma = matrix(c(1, 0.3, 0.3, 1), 2))
  )
plot(dataset)
do <- dist(dataset)
dbscan::kNNdist(do, 4) %>% apply(1, median) %>% summary()
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#> 0.02362 0.07265 0.10334 0.16972 0.16889 1.99385 

ci <- condense(do, 0.103)
dataset[ci, ] %>% plot()
dataset[condense(do, 0.68), ] %>% plot()