Removed LightElkan and updated MLJ

Author: Andrey Oskin

Project.toml

@@ -1,7 +1,7 @@
 name = "ParallelKMeans"
 uuid = "42b8e9d4-006b-409a-8472-7f34b3fb58af"
 authors = ["Bernard Brenyah", "Andrey Oskin"]
-version = "0.1.1"
+version = "0.1.2"
 
 [deps]
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"

src/ParallelKMeans.jl

@@ -7,13 +7,12 @@ import Distances
 
 include("seeding.jl")
 include("kmeans.jl")
-include("light_elkan.jl")
 include("lloyd.jl")
 include("hamerly.jl")
 include("elkan.jl")
 include("mlj_interface.jl")
 
 export kmeans
-export Lloyd, LightElkan, Hamerly, Elkan
+export Lloyd, Hamerly, Elkan
 
 end # module

src/elkan.jl

@@ -29,11 +29,12 @@ function kmeans!(alg::Elkan, containers, X, k;
     @parallelize n_threads ncol chunk_initialize(alg, containers, centroids, X)
 
     converged = false
-    niters = 1
+    niters = 0
     J_previous = 0.0
 
     # Update centroids & labels with closest members until convergence
-    while niters <= max_iters
+    while niters < max_iters
+        niters += 1
         # Core iteration
         @parallelize n_threads ncol chunk_update_centroids(alg, containers, centroids, X)
 
@@ -66,7 +67,6 @@ function kmeans!(alg::Elkan, containers, X, k;
         # Step 1 in original paper, calulation of distance d(c, c')
         update_containers(alg, containers, centroids, n_threads)
         J_previous = J
-        niters += 1
     end
 
     @parallelize n_threads ncol sum_of_squares(containers, X, containers.labels, centroids)

src/hamerly.jl

@@ -57,7 +57,6 @@ function kmeans!(alg::Hamerly, containers, X, k;
         end
 
         J_previous = J
-
     end
 
     @parallelize n_threads ncol sum_of_squares(containers, X, containers.labels, centroids)

src/light_elkan.jl

@@ -125,7 +125,7 @@ function chunk_update_centroids(::Lloyd, containers, centroids, X, r, idx)
     containers.J[idx] = J
 end
 
-function collect_containers(alg::T, containers, centroids, n_threads) where {T <: Union{LightElkan, Lloyd}}
+function collect_containers(alg::Lloyd, containers, centroids, n_threads)
     if n_threads == 1
         @inbounds centroids .= containers.centroids_new[1] ./ containers.centroids_cnt[1]'
     else

src/lloyd.jl

@@ -5,7 +5,7 @@ const ParallelKMeans_Desc = "Parallel & lightning fast implementation of all ava
 # availalbe variants for reference
 const MLJDICT = Dict(:Lloyd => Lloyd(),
                      :Hamerly => Hamerly(),
-                     :LightElkan => LightElkan())
+                     :Elkan => Elkan())
 
 ####
 #### MODEL DEFINITION
@@ -24,7 +24,7 @@ mutable struct KMeans <: MLJModelInterface.Unsupervised
 end
 
 
-function KMeans(; algo=:Lloyd, k_init="k-means++",
+function KMeans(; algo=:Hamerly, k_init="k-means++",
                 k=3, tol=1e-6, max_iters=300, copy=true,
                 threads=Threads.nthreads(), verbosity=0, init=nothing)
 
@@ -39,12 +39,12 @@ function MLJModelInterface.clean!(m::KMeans)
     warning = ""
 
     if !(m.algo ∈ keys(MLJDICT))
-        warning *= "Unsupported KMeans variant, Defauting to KMeans++ seeding algorithm."
-        m.algo = :Lloyd
+        warning *= "Unsupported KMeans variant, Defaulting to Hamerly algorithm."
+        m.algo = :Hamerly
 
     elseif m.k_init != "k-means++"
-        warning *= "Only `k-means++` or random seeding algorithms are supported. Defaulting to random seeding."
-        m.k_init = "random"
+        warning *= "Only `k-means++` or random seeding algorithms are supported. Defaulting to k-means++ seeding."
+        m.k_init = "kmeans++"
 
     elseif m.k < 1
         warning *= "Number of clusters must be greater than 0. Defaulting to 3 clusters."
@@ -63,8 +63,8 @@ function MLJModelInterface.clean!(m::KMeans)
         m.threads = Threads.nthreads()
 
     elseif !(m.verbosity ∈ (0, 1))
-        warning *= "Verbosity must be either 0 (no info) or 1 (info requested). Defaulting to 0."
-        m.verbosity = 0
+        warning *= "Verbosity must be either 0 (no info) or 1 (info requested). Defaulting to 1."
+        m.verbosity = 1
     end
     return warning
 end

src/mlj_interface.jl

@@ -1,69 +1,9 @@
 module TestElkan
 
 using ParallelKMeans
-using ParallelKMeans: update_containers
 using Test
 using Random
 
-@testset "centroid distances" begin
-    containers = (centroids_dist = Matrix{Float64}(undef, 3, 3), )
-    centroids = [1.0 2.0 4.0; 2.0 1.0 3.0]
-    update_containers(LightElkan(), containers, centroids, 1)
-    centroids_dist = containers.centroids_dist
-    @test centroids_dist[1, 2] == centroids_dist[2, 1]
-    @test centroids_dist[1, 3] == centroids_dist[3, 1]
-    @test centroids_dist[2, 3] == centroids_dist[3, 2]
-    @test centroids_dist[1, 2] == 0.5
-    @test centroids_dist[1, 3] == 2.5
-    @test centroids_dist[2, 3] == 2.0
-    @test centroids_dist[1, 1] == 0.5
-    @test centroids_dist[2, 2] == 0.5
-    @test centroids_dist[3, 3] == 2.0
-end
-
-@testset "basic kmeans light elkan" begin
-    X = [1. 2. 4.;]
-    res = kmeans(LightElkan(), X, 1; n_threads = 1, tol = 1e-6, verbose = false)
-    @test res.assignments == [1, 1, 1]
-    @test res.centers[1] ≈ 2.3333333333333335
-    @test res.totalcost ≈ 4.666666666666666
-    @test res.converged
-
-    res = kmeans(LightElkan(), X, 2; n_threads = 1, init = [1.0 4.0], tol = 1e-6, verbose = false)
-    @test res.assignments == [1, 1, 2]
-    @test res.centers ≈ [1.5 4.0]
-    @test res.totalcost ≈ 0.5
-    @test res.converged
-end
-
-@testset "no convergence yield last result" begin
-    X = [1. 2. 4.;]
-    res = kmeans(LightElkan(), X, 2; n_threads = 1, init = [1.0 4.0], tol = 1e-6, max_iters = 1, verbose = false)
-    @test !res.converged
-    @test res.totalcost ≈ 0.5
-end
-
-@testset "singlethread linear separation" begin
-    Random.seed!(2020)
-
-    X = rand(3, 100)
-    res = kmeans(LightElkan(), X, 3; n_threads = 1, tol = 1e-6, verbose = false)
-
-    @test res.totalcost ≈ 14.16198704459199
-    @test res.converged
-    @test res.iterations == 11
-end
-
-@testset "multithread linear separation quasi two threads" begin
-    Random.seed!(2020)
-
-    X = rand(3, 100)
-    res = kmeans(LightElkan(), X, 3; n_threads = 2, tol = 1e-6, verbose = false)
-
-    @test res.totalcost ≈ 14.16198704459199
-    @test res.converged
-end
-
 @testset "basic kmeans elkan" begin
     X = [1. 2. 4.;]
     res = kmeans(Elkan(), X, 1; n_threads = 1, tol = 1e-6, verbose = false)
@@ -94,7 +34,7 @@ end
 
     @test res.totalcost ≈ 14.16198704459199
     @test !res.converged
-    @test res.iterations == 11
+    @test res.iterations == 10
 end
 
 @testset "elkan multithread linear separation quasi two threads" begin