WIP: Faster metric support for Lloyd

Author: PyDataBlog

Project.toml

@@ -8,6 +8,7 @@ Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
 MLJModelInterface = "e80e1ace-859a-464e-9ed9-23947d8ae3ea"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
+UnsafeArrays = "c4a57d5a-5b31-53a6-b365-19f8c011fbd6"
 
 [compat]
 Distances = "0.8.2"

src/ParallelKMeans.jl

@@ -2,6 +2,7 @@ module ParallelKMeans
 
 using StatsBase
 using Random
+using UnsafeArrays
 import MLJModelInterface
 import Base.Threads: @spawn
 import Distances

src/kmeans.jl

@@ -102,7 +102,7 @@ end
 
 Allocationless calculation of distance between vectors X1[:, i1] and X2[:, i2] defined by the supplied distance metric.
 """
-distance(metric, X1, X2, i1, i2) = evaluate(metric, X1[:, i1], X2[:, i2])
+@inline distance(metric, X1, X2, i1, i2) = evaluate(metric, uview(X1, :, i1), uview(X2, :, i2))
 
 
 """

test/test03_lloyd.jl

@@ -118,7 +118,7 @@ end
     X = rand(3, 100)
 
     res = kmeans(X, 2, tol = 1e-16, metric=Cityblock())
-    @test res.totalcost ≈ 62.040452528953736
+    @test res.totalcost ≈ 62.04045252895372
     @test res.converged
 end