Nearest neighbors and distances

Obtain nearest neighbors and distances from a matrix or disto handle. k nearest or fixed radius neighbors are supported

nn(x, k, r, ...)

Arguments

x	Object of class 'disto' or a numeric matrix
k	Number of nearest neighbors
r	Radius for nearest neighbors
...	Additional arguments for dapply when x is 'disto' object. Else additional arguments are sent to pbmclapply

Value

A list with these elements:

• triplet: Matrix with three columns: row, col and distance. For a fixed observation(value in 'row'), all corresponding values in 'col' are the indexes of the nearest neighbors. All corresponding values in 'distance' are the distances to those nearest neighbors

• size: Size of the distance matrix or number of rows of the matrix

• k or r : Depending on the input

Details

Exactly one among k or r has to be provided

Examples

# NOT RUN {
# create a matrix
set.seed(100)
mat <- cbind(rnorm(3e3), rpois(3e3, 1))

# compute a distance matrix and get a disto handle
do <- stats::dist(mat)
dio <- disto(objectname = "do")

# nearest neighbors: k nearest and fixed radius
nn(dio, k = 1)
nn(mat, k = 1) # distance method defaults to 'euclidean'
str(nn(mat, k = 1)) # observe the structure of the output

nn(dio, r = 0.1)
nn(mat, r = 0.1)

# nearest neighbors parallelized: k nearest and fixed radius
# fast computation, higher memory usage
nn(dio, k = 1, nproc = 2)
nn(mat, k = 1, mc.cores = 2)

nn(dio, r = 0.1, nproc = 2)
nn(mat, r = 0.1, mc.cores = 2)

# different distance method
do <- stats::dist(mat, method = "manhattan")

nn(dio, k = 1, nproc = 2)
nn(mat, k = 1, method = "manhattan", mc.cores = 2)

nn(dio, r = 0.1, nproc = 2)
nn(mat, r = 0.1, method = "manhattan", mc.cores = 2)
# }