Basic example for NN

[rveltz]

Hi,

Is it possible to provide a basic example to search nearest neighbors for say data = rand(1000, 3)?

Thank you

[efaulhaber]

More extensive documentation is on my to-do list. Here is a basic example:

using PointNeighbors

nhs = GridNeighborhoodSearch{3}(search_radius = 0.1)
# Initialize the NHS to find neighbors in x of particles in x
initialize!(nhs, x, x)

# Simple example: just count the neighbors of each particle
n_neighbors = zeros(Int, 1000);

foreach_point_neighbor(x, x, nhs) do i, j, pos_diff, distance
    n_neighbors[i] += 1
end

This results in:

julia> n_neighbors
1000-element Vector{Int64}:
  3
  9
 10
  8
  5
  8
  8
  4
  6
  3
  2
  1
 10
  3
  5
  ⋮
  2
  4
  4
  2
  5
  7
  5
  4
  5
  4
  5
 10
  6
 12

Note that this is not n nearest neighbors, but we find all neighbors within a fixed search radius.

[rveltz]

I just want the nearest neightbor of a random point. Is it possible? (I want to test your package versus NearestNeighbors.jl which "does not work on gpu"

[efaulhaber]

No, this is not possible. Unless you know a good upper bound on the maximum distance between nearest points. Then you could use this as search radius and then compute the minimum over the neighbors within the search radius.