Coverage for GNFA and similarity methods

Project.toml

@@ -5,7 +5,6 @@ description = "A Julia package for Adaptive Resonance Theory (ART) algorithms."
 version = "0.1.0"
 
 [deps]
-DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Logging = "56ddb016-857b-54e1-b83d-db4d58db5568"
 MLJ = "add582a8-e3ab-11e8-2d5e-e98b27df1bc7"
@@ -23,9 +22,10 @@ PyPlot = "2.9.0"
 julia = "1"
 
 [extras]
+DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"
 MLDatasets = "eb30cadb-4394-5ae3-aed4-317e484a6458"
 SafeTestsets = "1bc83da4-3b8d-516f-aca4-4fe02f6d838f"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Test", "SafeTestsets", "MLDatasets"]
+test = ["Test", "SafeTestsets", "MLDatasets", "DelimitedFiles"]

src/AdaptiveResonance.jl

@@ -1,10 +1,5 @@
 module AdaptiveResonance
 
-# using Logging
-# using Parameters
-# using Statistics
-# using LinearAlgebra
-
 abstract type AbstractARTOpts end
 abstract type AbstractART end
 

src/DDVFA.jl

@@ -5,7 +5,6 @@ using LinearAlgebra
 using ProgressBars
 using Printf
 
-
 """
     opts_GNFA()
 
@@ -39,7 +38,6 @@ Initialized GNFA
     max_epochs = 1
 end # opts_GNFA
 
-
 """
     GNFA
 
@@ -72,7 +70,6 @@ mutable struct GNFA <: AbstractART
     epoch::Int
 end # GNFA
 
-
 """
     GNFA()
 
@@ -91,7 +88,6 @@ function GNFA()
     GNFA(opts)
 end # GNFA()
 
-
 """
     GNFA(opts)
 
@@ -121,7 +117,6 @@ function GNFA(opts)
     )
 end # GNFA(opts)
 
-
 """
     initialize!()
 
@@ -150,7 +145,6 @@ function initialize!(art::GNFA, x::Array)
     # push!(art.labels, label)
 end # initialize!(GNFA, x)
 
-
 """
     train!()
 
@@ -309,7 +303,6 @@ function classify(art::GNFA, x::Array)
     return y_hat
 end # classify(GNFA, x)
 
-
 """
     activation_match!(art::GNFA, x::Array)
 
@@ -339,25 +332,27 @@ end # activation_match!(GNFA, x)
 
 
 # Generic learning function
-function learn(art::GNFA, x, W)
+function learn(art::GNFA, x::Array, W::Array)
     # Update W
     return art.opts.beta .* element_min(x, W) .+ W .* (1 - art.opts.beta)
 end # learn(GNFA, x, W)
 
-
 # In place learning function with instance counting
-function learn!(art::GNFA, x, index)
+function learn!(art::GNFA, x::Array, index::Int)
     # Update W
     art.W[:, index] = learn(art, x, art.W[:, index])
     art.n_instance[index] += 1
 end # learn!(GNFA, x, index)
 
+"""
+    stopping_conditions(art::GNFA)
 
+Stopping conditions for a GNFA module.
+"""
 function stopping_conditions(art::GNFA)
     return isequal(art.W, art.W_old) || art.epoch >= art.opts.max_epochs
 end # stopping_conditions(GNFA)
 
-
 """
     opts_DDVFA()
 
@@ -449,7 +444,6 @@ function DDVFA(opts::opts_DDVFA)
     )
 end # DDVFA(opts)
 
-
 """
     train!(ddvfa, data)
 
@@ -538,7 +532,6 @@ function train!(art::DDVFA, x::Array)
     end
 end # train!(DDVFA, x)
 
-
 """
     stopping_conditions(art::DDVFA)
 
@@ -551,7 +544,6 @@ function stopping_conditions(art::DDVFA)
     return art.W == art.W_old || art.epoch >= art.opts.max_epoch
 end # stopping_conditions(DDVFA)
 
-
 """
     similarity(method, F2, field_name, gamma_ref)
 
@@ -595,14 +587,15 @@ function similarity(method::String, F2::GNFA, field_name::String, sample::Array,
     # Weighted linkage
     elseif method == "weighted"
         if field_name == "T"
-            value = F2.T * (F2.n / sum(F2.n))
+            value = F2.T' * (F2.n_instance ./ sum(F2.n_instance))
         elseif field_name == "M"
-            value = F2.M * (F2.n / sum(F2.n))
+            value = F2.M' * (F2.n_instance ./ sum(F2.n_instance))
         end
     # Centroid linkage
     elseif method == "centroid"
-        Wc = minimum(F2.W)
-        T = norm(min(sample, Wc), 1)
+        Wc = minimum(F2.W, dims=2)
+        # (norm(min(obj.sample, Wc), 1)/(obj.alpha + norm(Wc, 1)))^obj.gamma;
+        T = norm(element_min(sample, Wc), 1) / (F2.opts.alpha + norm(Wc, 1))^F2.opts.gamma
         if field_name == "T"
             value = T
         elseif field_name == "M"

src/funcs.jl

@@ -21,6 +21,11 @@ function complement_code(data::Array)
     return x
 end
 
+"""
+    element_min(x::Array, W::Array)
+
+Returns the element-wise minimum between sample x and weight W.
+"""
 function element_min(x::Array, W::Array)
     # Compute the element-wise minimum of two vectors
     return minimum([x W], dims = 2)
@@ -152,4 +157,3 @@ end
 #         error("Invalid/unimplemented similarity method")
 #     end
 # end # similarity
-

test/test_sets.jl

@@ -25,7 +25,35 @@ end
 
     # GNFA train and test
     my_gnfa = GNFA()
-
+    data = load_am_data(200, 50)
+    local_complement_code = AdaptiveResonance.complement_code(data.train_x)
+    train!(my_gnfa, local_complement_code)
+
+    # Similarity methods
+    methods = ["single",
+               "average",
+               "complete",
+               "median",
+               "weighted",
+               "centroid"]
+
+    # Both field names
+    field_names = ["T", "M"]
+
+    # Compute a local sample for GNFA similarity method testing
+    local_sample = local_complement_code[:, 1]
+
+    # Compute the local activation and match
+    AdaptiveResonance.activation_match!(my_gnfa, local_sample)
+
+    # Test every method and field name
+    for method in methods
+        println("Method: ", method)
+        for field_name in field_names
+            result = AdaptiveResonance.similarity(method, my_gnfa, field_name, local_sample, my_gnfa.opts.gamma_ref)
+            println(field_name, ": ", result)
+        end
+    end
 end
 
 

test/test_utils.jl

@@ -22,10 +22,13 @@ dataset and packages them into a DataSplit struct.
 function load_am_data(N_train::Int, N_test::Int)
     # Load the data, downloading if in a CI context: TODO
     # if ENV["CI"] == true
-    MNIST.download("../data/mnist/", i_accept_the_terms_of_use=true)
+    data_dir = "../data/mnist/"
+    if !isdir(data_dir)
+        MNIST.download(data_dir, i_accept_the_terms_of_use=true)
+    end
     # end
-    train_x, train_y = MNIST.traindata()
-    test_x, test_y = MNIST.testdata()
+    train_x, train_y = MNIST.traindata(dir=data_dir)
+    test_x, test_y = MNIST.testdata(dir=data_dir)
 
     # Get sizes of train and test data
     size_a, size_b, data_n = size(train_x)