Merge pull request #409 from JuliaTrustworthyAI/408-get-coverage-up-t…

…o-80

increase for data-processing

Project.toml

@@ -27,7 +27,6 @@ Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
-cuDNN = "02a925ec-e4fe-4b08-9a7e-0d78e3d38ccd"
 
 [weakdeps]
 EvoTrees = "f6006082-12f8-11e9-0c9c-0d5d367ab1e5"
@@ -38,10 +37,9 @@ EvoTreesExt = "EvoTrees"
 LaplaceReduxExt = "LaplaceRedux"
 
 [compat]
+CUDA = "3, 4, 5"
 CategoricalArrays = "0.10"
 ChainRulesCore = "1.15"
-CUDA = "3, 4, 5"
-cuDNN = "1"
 DataFrames = "1"
 DecisionTree = "0.12.3"
 Distributions = "0.25.97"

docs/setup_docs.jl

@@ -36,7 +36,7 @@ setup_docs = quote
     ENV["DATADEPS_ALWAYS_ACCEPT"] = "true"
 
     # Counteractual data and model:
-    counterfactual_data = load_linearly_separable()
+    counterfactual_data = CounterfactualData(load_linearly_separable()...)
     M = fit_model(counterfactual_data, :Linear)
     target = 2
     factual = 1

src/data_preprocessing/counterfactual_data.jl

@@ -1,3 +1,5 @@
+using MLJBase
+
 """
     CounterfactualData(
         X::AbstractMatrix, y::AbstractMatrix;
@@ -181,10 +183,14 @@ Outer constructor method that accepts a `Tables.MatrixTable`. By default, the in
 
 """
 function CounterfactualData(X::Tables.MatrixTable, y::RawOutputArrayType; kwrgs...)
-    features_categorical = findall([scitype(x) <: AbstractVector{<:Finite} for x in X])
+    features_categorical = findall([
+        MLJBase.scitype(x) <: AbstractVector{<:Finite} for x in X
+    ])
     features_categorical =
         length(features_categorical) == 0 ? nothing : features_categorical
-    features_continuous = findall([scitype(x) <: AbstractVector{<:Continuous} for x in X])
+    features_continuous = findall([
+        MLJBase.scitype(x) <: AbstractVector{<:Continuous} for x in X
+    ])
     features_continuous = length(features_continuous) == 0 ? nothing : features_continuous
     X = permutedims(Tables.matrix(X))
 

test/Project.toml

@@ -1,4 +1,5 @@
 [deps]
+Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
 Chain = "8be319e6-bccf-4806-a6f7-6fae938471bc"
 CompatHelperLocal = "5224ae11-6099-4aaa-941d-3aab004bd678"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
@@ -7,16 +8,16 @@ Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
 LaplaceRedux = "c52c1a26-f7c5-402b-80be-ba1e638ad478"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Logging = "56ddb016-857b-54e1-b83d-db4d58db5568"
+MLDatasets = "eb30cadb-4394-5ae3-aed4-317e484a6458"
 MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MLJDecisionTreeInterface = "c6f25543-311c-4c74-83dc-3ea6d1015661"
+MLJModels = "d491faf4-2d78-11e9-2867-c94bc002c0b7"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
-Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 TaijaData = "9d524318-b4e6-4a65-86d2-b2b72d07866c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
-cuDNN = "02a925ec-e4fe-4b08-9a7e-0d78e3d38ccd"
-Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
 
 [compat]
-Aqua = "0.8"
+Aqua = "0.8"

test/data/data_preprocessing.jl

@@ -25,3 +25,13 @@ ENV["DATADEPS_ALWAYS_ACCEPT"] = "true"
     counterfactual_data = DataPreprocessing.CounterfactualData(X, y; domain=(0, 0))
     @test unique(DataPreprocessing.apply_domain_constraints(counterfactual_data, x))[1] == 0
 end
+
+@testset "Other" begin
+    # MatrixTable:
+    dataset = Iris()
+    X = dataset.features
+    y = dataset.targets
+    X = MLJBase.table(Tables.matrix(X))
+    y = y[:, 1]
+    counterfactual_data = CounterfactualData(X, y)
+end

test/generators/categorical_data.jl

@@ -0,0 +1,50 @@
+using MLJModels: OneHotEncoder
+using StatsBase: sample, Weights
+
+@testset "Categorical data" begin
+
+    # Set up a simple categorical dataset:
+    N = 1000
+    X, ys = MLJBase.make_blobs(
+        N, 2; centers=2, as_table=false, center_box=(-5 => 5), cluster_std=0.5
+    )
+    ys .= ys .== 2
+    cat_values = ["X", "Y", "Z"]
+    xcat = map(ys) do y
+        if y == 1
+            x = sample(cat_values, Weights([0.8, 0.1, 0.1]))
+        else
+            x = sample(cat_values, Weights([0.1, 0.1, 0.8]))
+        end
+    end
+    xcat = categorical(xcat)
+    X = (x1=X[:, 1], x2=X[:, 2], x3=xcat)
+
+    # Encode the categorical feature:
+    hot = OneHotEncoder()
+    mach = MLJBase.fit!(machine(hot, X))
+    W = MLJBase.transform(mach, X)
+    X = permutedims(MLJBase.matrix(W))
+
+    # Create a counterfactual data object:
+    features_categorical = [collect(3:size(X, 1))]
+    counterfactual_data = CounterfactualData(
+        X, ys'; features_categorical=features_categorical
+    )
+
+    # Fit a model:
+    M = fit_model(counterfactual_data, :Linear)
+
+    # Sample a factual:
+    target = 1
+    factual = 0
+    chosen = rand(findall(predict_label(M, counterfactual_data) .== factual))
+    x = select_factual(counterfactual_data, chosen)
+
+    # Generate a counterfactual:
+    generator = GenericGenerator()
+    ce = generate_counterfactual(x, target, counterfactual_data, M, generator)
+
+    # Test:
+    @test true
+end

test/generators/generators.jl

@@ -3,6 +3,7 @@ include("feature_tweak.jl")
 include("counterfactuals.jl")
 include("probe.jl")
 include("growing_spheres.jl")
+include("categorical_data.jl")
 
 # generic tests
 @testset "Construction" begin

test/other/objectives.jl

@@ -0,0 +1,27 @@
+using CounterfactualExplanations
+using CounterfactualExplanations.Models: fit_model
+using CounterfactualExplanations.Objectives
+
+# Counteractual data and model:
+counterfactual_data = CounterfactualData(TaijaData.load_linearly_separable()...)
+M = fit_model(counterfactual_data, :Linear)
+target = 2
+factual = 1
+chosen = rand(findall(predict_label(M, counterfactual_data) .== factual))
+x = select_factual(counterfactual_data, chosen)
+
+# Search:
+generator = GenericGenerator()
+ce = generate_counterfactual(x, target, counterfactual_data, M, generator)
+
+@testset "Losses" begin
+    for (lname, lfun) in Objectives.losses_catalogue
+        @test lfun(ce) isa AbstractFloat
+    end
+end
+
+@testset "Penalties" begin
+    for (pname, pfun) in Objectives.penalties_catalogue
+        @test pfun(ce) isa AbstractFloat
+    end
+end

test/other/other.jl

@@ -0,0 +1,3 @@
+CounterfactualExplanations.reset!(flux_training_params)
+
+@test true

test/parallelization/parallelization.jl

@@ -1,3 +1,9 @@
+using CounterfactualExplanations
+
+@test parallelizable(generate_counterfactual) == true
+@test parallelizable(CounterfactualExplanations.Evaluation.evaluate) == true
+@test parallelizable(predict_label) == false
+
 @testset "Threads" begin
     include("threads.jl")
 end

test/parallelization/threads.jl

@@ -1,4 +1,5 @@
 using CounterfactualExplanations
+using CounterfactualExplanations.Evaluation: evaluate
 using CounterfactualExplanations.Parallelization
 
 counterfactual_data = synthetic[:classification_binary][:data]
@@ -14,4 +15,6 @@ ces = @with_parallelizer parallelizer begin
     generate_counterfactual(xs, target, counterfactual_data, M, generator)
 end
 
+@with_parallelizer parallelizer evaluate(ces)
+
 @test true

test/runtests.jl

@@ -1,23 +1,23 @@
 using Chain: @chain
 using CounterfactualExplanations
+using CounterfactualExplanations.Convergence
+using CounterfactualExplanations.DataPreprocessing
+using CounterfactualExplanations.Evaluation
+using CounterfactualExplanations.Generators
+using CounterfactualExplanations.Models
 using Test
 using DataFrames
+using EvoTrees
 using Flux
-using LinearAlgebra
-using MLUtils
-using Random
-using Plots
 using LaplaceRedux
-using EvoTrees
+using LinearAlgebra
+using MLDatasets
 using MLJBase
 using MLJDecisionTreeInterface
 using Printf
-using CounterfactualExplanations.Convergence
+using MLUtils
+using Random
 using TaijaData
-using CounterfactualExplanations.Generators
-using CounterfactualExplanations.Models
-using CounterfactualExplanations.Evaluation
-using CounterfactualExplanations.DataPreprocessing
 
 Random.seed!(0)
 
@@ -52,7 +52,15 @@ generators = Generators.generator_catalogue
         include("other/evaluation.jl")
     end
 
+    @testset "Objectives" begin
+        include("other/objectives.jl")
+    end
+
     @testset "Parallelization" begin
         include("parallelization/parallelization.jl")
     end
+
+    @testset "Other" begin
+        include("other/other.jl")
+    end
 end