Added verbosity to MLJ interface fit

Author: PyDataBlog

src/mlj_interface.jl

@@ -19,16 +19,15 @@ mutable struct KMeans <: MMI.Unsupervised
     max_iters::Int
     copy::Bool
     threads::Int
-    verbosity::Int
     init
 end
 
 
 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)
+                threads=Threads.nthreads(), init=nothing)
 
-    model   = KMeans(algo, k_init, k, tol, max_iters, copy, threads, verbosity, init)
+    model   = KMeans(algo, k_init, k, tol, max_iters, copy, threads, init)
     message = MMI.clean!(model)
     isempty(message) || @warn message
     return model
@@ -68,11 +67,6 @@ function MMI.clean!(m::KMeans)
         m.threads = Threads.nthreads()
 	end
 
-    if !(m.verbosity ∈ (0, 1))
-        push!(warning, "Verbosity must be either 0 (no info) or 1 (info requested). Defaulting to 1.")
-        m.verbosity = 1
-    end
-
     return join(warning, "\n")
 end
 
@@ -85,7 +79,7 @@ end
 
     See also the [package documentation](https://pydatablog.github.io/ParallelKMeans.jl/stable).
 """
-function MMI.fit(m::KMeans, X)
+function MMI.fit(m::KMeans, verbosity::Int, X)
     # convert tabular input data into the matrix model expects. Column assumed as features so input data is permuted
     if !m.copy
         # permutes dimensions of input table without copying and pass to model
@@ -99,16 +93,22 @@ function MMI.fit(m::KMeans, X)
     algo = MLJDICT[m.algo]  # select algo
 
     # fit model and get results
-    verbose = m.verbosity != 0
+    verbose = verbosity > 0  # Display fitting operations if verbosity > 0
     fitresult = ParallelKMeans.kmeans(algo, DMatrix, m.k;
                                       n_threads = m.threads, k_init=m.k_init,
                                       max_iters=m.max_iters, tol=m.tol, init=m.init,
                                       verbose=verbose)
+
     cache = nothing
     report = (cluster_centers=fitresult.centers, iterations=fitresult.iterations,
               converged=fitresult.converged, totalcost=fitresult.totalcost,
               labels=fitresult.assignments)
-
+    """
+    # TODO: warn users about non convergence
+    if verbose & (!fitresult.converged)
+        @warn "Specified model failed to converge."
+    end
+    """
     return (fitresult, cache, report)
 end
 
@@ -144,7 +144,7 @@ function MMI.transform(m::KMeans, fitresult, Xnew)
 
     # Warn users if fitresult is from a `non-converged` fit
     if !fitresult[end].converged
-        @warn "Failed to converged. Using last assignments to make transformations."
+        @warn "Failed to converge. Using last assignments to make transformations."
     end
 
     # results from fitted model
@@ -175,7 +175,7 @@ MMI.metadata_pkg.(KMeans,
 # Metadata for ParaKMeans model interface
 MMI.metadata_model(KMeans,
     input   = MMI.Table(MMI.Continuous),
-    output  = MMI.Table(MMI.Count),
+    output  = MMI.Table(MMI.Continuous),
     weights = false,
     descr   = ParallelKMeans_Desc,
 	path	= "ParallelKMeans.KMeans")

test/test07_mlj_interface.jl

@@ -19,7 +19,6 @@ using MLJBase
     @test model.copy            == true
     @test model.threads         == Threads.nthreads()
     @test model.tol             == 1.0e-6
-    @test model.verbosity       == 0
 end
 
 
@@ -30,15 +29,14 @@ end
     @test_logs (:warn, "Tolerance level must be less than 1. Defaulting to tol of 1e-6.") ParallelKMeans.KMeans(tol=2)
     @test_logs (:warn, "Number of permitted iterations must be greater than 0. Defaulting to 300 iterations.") ParallelKMeans.KMeans(max_iters=0)
     @test_logs (:warn, "Number of threads must be at least 1. Defaulting to all threads available.") ParallelKMeans.KMeans(threads=0)
-    @test_logs (:warn, "Verbosity must be either 0 (no info) or 1 (info requested). Defaulting to 1.") ParallelKMeans.KMeans(verbosity=100)
 end
 
 
 @testset "Test model fitting verbosity" begin
     Random.seed!(2020)
     X = table([1 2; 1 4; 1 0; 10 2; 10 4; 10 0])
-    model = KMeans(k=2, max_iters=1, verbosity=1)
-    results = @capture_out fit(model, X)
+    model = KMeans(k=2, max_iters=1)
+    results = @capture_out fit(model, 1, X)
 
     @test results == "Iteration 1: Jclust = 28.0\n"
 end
@@ -50,7 +48,7 @@ end
     X_test = table([10 1])
 
     model = KMeans(algo = :Lloyd, k=2)
-    results = fit(model, X)
+    results = fit(model, 0, X)
 
     @test results[2]             == nothing
     @test results[end].converged == true
@@ -72,7 +70,7 @@ end
     X_test = table([10 1])
 
     model = KMeans(algo=:Hamerly, k=2)
-    results = fit(model, X)
+    results = fit(model, 0, X)
 
     @test results[2]             == nothing
     @test results[end].converged == true
@@ -87,13 +85,14 @@ end
     @test preds[:x1][1] == 2
 end
 
+
 @testset "Test Elkan model fitting" begin
     Random.seed!(2020)
     X = table([1 2; 1 4; 1 0; 10 2; 10 4; 10 0])
     X_test = table([10 1])
 
     model = KMeans(algo=:Elkan, k=2)
-    results = fit(model, X)
+    results = fit(model, 0, X)
 
     @test results[2]             == nothing
     @test results[end].converged == true
@@ -108,15 +107,26 @@ end
     @test preds[:x1][1] == 2
 end
 
+
 @testset "Testing non convergence warning" begin
     Random.seed!(2020)
     X = table([1 2; 1 4; 1 0; 10 2; 10 4; 10 0])
     X_test = table([10 1])
 
     model = KMeans(k=2, max_iters=1)
-    results = fit(model, X)
+    results = fit(model, 0, X)
 
-    @test_logs (:warn, "Failed to converged. Using last assignments to make transformations.") transform(model, results, X_test)
+    @test_logs (:warn, "Failed to converge. Using last assignments to make transformations.") transform(model, results, X_test)
 end
 
+"""
+@testset "Testing non convergence warning during model fitting" begin
+    Random.seed!(2020)
+    X = table([1 2; 1 4; 1 0; 10 2; 10 4; 10 0])
+    X_test = table([10 1])
+
+    model = KMeans(k=2, max_iters=1)
+    @test_logs (:warn, "Specified model failed to converge.") fit(model, 1, X);
+end
+"""
 end # module