Clustering Functions
#
DPMeansClustering.dp_means — Function.
dp_means(input_vectors::Array{Vector{Int}, 1}, radius::Float64; verbose = false)
Cluster input_vectors using euclidean distance metric and arithmetic mean, where a vector with distance greater than radius from the nearest cluster forms a new cluster.
Returns (centroid_vectors, cluster_sizes, cluster_indices, centroid_indices) where cluster_indices is an array of arrays of indices of input_vectors grouped by cluster, and centroid_indices is an array of values where each value indexes the vector in input_vectors closest to the respective centroid.
Example:
inputs = [[5, 0, 0, 1, 1, 5],
[5, 1, 0, 0, 1, 5],
[5, 0, 1, 0, 1, 5],
[0, 4, 6, 2, 0, 0],
[0, 4, 6, 1, 1, 0],
[0, 4, 6, 1, 0, 1]]
radius = 3.0
μs, sizes, indices, centroids = dp_means(inputs, radius)
μs == [[5, 1/3, 1/3, 1/3, 1, 5],
[0, 4, 6, 4/3, 1/3, 1/3]]
sizes == [3, 3]
indices == [[1, 2, 3], [4, 5, 6]]
centroids == [1, 4]
#
DPMeansClustering.dp_centers — Function.
dp_centers(inputs, radius::Float64; distfunc = euclidean, center = mean, verbose = false, cycle_lim = 30)
Cluster input_vectors using given distance and mean calculations, where a vector with distance greater than radius from the nearest cluster forms a new cluster. Runs a maximum of cycle_lim iterations.
Returns (centroid_vectors, cluster_sizes, cluster_indices, centroid_indices) where cluster_indices is an array of arrays of indices of input_vectors grouped by cluster, and centroid_indices is an array of values where each value indexes the vector in input_vectors closest to the respective centroid.
This is a generalization of dp_means.
dp_centers(inputs, radii::Vector{Float64}; distfunc = euclidean, center = mean, verbose = false, cycle_lim = [30...])
Cluster input_vectors using given distance and mean calculations recursively; radii should be an array of decreasing values, where each value is the radius of a successively finer clustering operation. During each clustering, a vector with distance greater than radius from the nearest cluster forms a new cluster. Runs a maximum of cycle_lims iterations during each cluster.
Returns (centroid_vectors, cluster_sizes, cluster_indices, centroid_indices) where cluster_indices is an array of arrays of indices of input_vectors grouped by cluster, and centroid_indices is an array of values where each value indexes the vector in input_vectors closest to the respective centroid.
This is a recursive implementation of dp_centers(inputs, radius) for faster clustering.
#
DPMeansClustering.cluster — Function.
cluster(inputs, radius::Float64; distfunc=euclidean, center=mean, verbose::Int64=1,
cycle_lim::Int64=30, triangle=false)
A wrapper that decides whether or not to use triangle inequality.