Remove MLJ dependency, redefine performance, bump

Project.toml

@@ -2,14 +2,13 @@ name = "AdaptiveResonance"
 uuid = "3d72adc0-63d3-4141-bf9b-84450dd0395b"
 authors = ["Sasha Petrenko"]
 description = "A Julia package for Adaptive Resonance Theory (ART) algorithms."
-version = "0.3.0"
+version = "0.3.1"
 
 [deps]
 ClusterValidityIndices = "2fefd308-f647-4698-a2f0-e90cfcbca9ad"
 Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Logging = "56ddb016-857b-54e1-b83d-db4d58db5568"
-MLJ = "add582a8-e3ab-11e8-2d5e-e98b27df1bc7"
 Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 ProgressBars = "49802e3a-d2f1-5c88-81d8-b72133a6f568"
@@ -18,9 +17,8 @@ Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
 [compat]
 ClusterValidityIndices = "0.1"
-MLJ = "0.14, 0.15, 0.16"
 Parameters = "0.12.1"
-ProgressBars = "0.7, 0.8, 1.0"
+ProgressBars = "0.7, 0.8, 1.0, 1.3"
 julia = "1"
 
 [extras]

src/ARTMAP/ARTSCENE.jl

@@ -178,7 +178,7 @@ function contrast_sensitive_oriented_filtering(image::RealArray, x::RealArray ;
         y[:,:,:,k] = deepcopy(x)
     end
 
-    for i = 1:n_iterations
+    for _ = 1:n_iterations
         y += ddt_y(y, X_plus, X_minus, alpha, distributed)
     end
 
@@ -223,7 +223,7 @@ end # competition_kernel(l::Integer, k::Integer ; sign::String="plus")
 
 Time rate of change for ARTSCENE: Stage 5.
 """
-function ddt_z(z::RealArray ; distributed=true)
+function ddt_z(z::RealArray ; distributed::Bool=true)
     n_row, n_column, n_g, n_k = size(z)
     kernel_r = 5
 
@@ -266,8 +266,7 @@ function orientation_competition(z::RealArray)
     #     Z[:,:,:,k] = deepcopy(z)
     # end
 
-    for i = 1:n_iterations
-        # Z += ddt_z(z)
+    for _ = 1:n_iterations
         z += ddt_z(z)
     end
 

src/AdaptiveResonance.jl

@@ -5,8 +5,7 @@ using Parameters    # ARTopts are parameters (@with_kw)
 using Logging       # Logging utils used as main method of terminal reporting
 using ProgressBars  # Provides progress bars for training and inference
 using Printf        # Used for printing formatted progress bars
-using MLJ: confusion_matrix, coerce, OrderedFactor
-using LinearAlgebra: tr, norm   # Trace and norms
+using LinearAlgebra: norm   # Trace and norms
 using Statistics: median, mean  # Medians and mean for linkage methods
 
 # Abstract types

src/common.jl

@@ -117,27 +117,23 @@ end # element_min(x::RealVector, W::RealVector)
 Returns the categorization performance of y_hat against y.
 """
 function performance(y_hat::IntegerVector, y::IntegerVector)
+    # Get the number of labels
+    n_y = length(y)
+
     # Check lengths
-    if length(y_hat) != length(y)
+    if length(y_hat) != n_y
         error("Label vectors must be the same length")
     end
 
-    # Clean up the vectors
-    n_mismatch = 0
-    y_hat_local = Integer[]
-    y_local = Integer[]
-    for ix = 1:length(y_hat)
-        if y_hat[ix] != -1
-            push!(y_hat_local, y_hat[ix])
-            push!(y_local, y[ix])
-        else
-            n_mismatch += 1
+    # Get the number of correct classifications
+    n_correct = 0
+    for ix = 1:n_y
+        if y_hat[ix] == y[ix]
+            n_correct += 1
         end
     end
 
-    # Compute the confusion matrix and calculate performance as trace/sum
-    conf = confusion_matrix(coerce(y_hat_local, OrderedFactor), coerce(y_local, OrderedFactor))
-    return tr(conf.mat)/(sum(conf.mat) + n_mismatch)
+    return n_correct/n_y
 end # performance(y_hat::IntegerVector, y::IntegerVector)
 
 """

test/test_ddvfa.jl

@@ -160,19 +160,33 @@ end # @testset "DDVFA"
     # Compute the local activation and match
     AdaptiveResonance.activation_match!(my_gnfa, local_sample)
 
-    truth = Dict("single" => Dict("T" => 0.9988445088278305,
-                                  "M" => 2.591300556893253),
-                 "average" => Dict("T" => 0.41577750468594143,
-                                   "M" => 1.322517210029363),
-                 "complete" => Dict("T" => 0.04556971777638373,
-                                    "M" => 0.13166315262229716),
-                 "median" => Dict("T" => 0.3312241307874298,
-                                  "M" => 1.3248965231497192),
-                 "weighted" => Dict("T" => 0.533208585217186,
-                                    "M" => 1.3855766656866793),
-                 "centroid" => Dict("T" => 0.0,
-                                    "M" => 0.0)
-                )
+    # Declare the true activation and match magnitudes
+    truth = Dict(
+        "single" => Dict(
+            "T" => 0.9988445088278305,
+            "M" => 2.591300556893253
+        ),
+        "average" => Dict(
+            "T" => 0.41577750468594143,
+            "M" => 1.322517210029363
+        ),
+        "complete" => Dict(
+            "T" => 0.04556971777638373,
+            "M" => 0.13166315262229716
+        ),
+        "median" => Dict(
+            "T" => 0.3312241307874298,
+            "M" => 1.3248965231497192
+        ),
+        "weighted" => Dict(
+            "T" => 0.533208585217186,
+            "M" => 1.3855766656866793
+        ),
+        "centroid" => Dict(
+            "T" => 0.0,
+            "M" => 0.0
+        )
+    )
 
     # Test every method and field name
     for method in methods