Merge pull request #434 from alan-turing-institute/dev

For a 0.15.3 release

Project.toml

@@ -1,7 +1,7 @@
 name = "MLJBase"
 uuid = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 authors = ["Anthony D. Blaom <[email protected]>"]
-version = "0.15.2"
+version = "0.15.3"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
@@ -32,7 +32,7 @@ Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 [compat]
 CategoricalArrays = "^0.8.1"
 ComputationalResources = "^0.3"
-Distributions = "^0.22,^0.23"
+Distributions = "0.22, 0.23, 0.24"
 HTTP = "^0.8"
 InvertedIndices = "^1"
 JLSO = "^2.1,^2.2"

src/composition/models/pipelines.jl

@@ -499,6 +499,11 @@ in which the specified models are composed in order. This means
 `model1` receives inputs, whose output is passed to `model2`, and so
 forth. Model types or instances may be specified.
 
+**Important.** By default a new model *type* name is automatically
+generated. To specify a different name add a keyword argument such as
+`name=MyPipeType`. This is necessary if serializing the pipeline; see
+[`MLJ.save`](@ref).
+
 At most one of the models may be a supervised model, but this model
 can appear in any position.
 

src/univariate_finite/methods.jl

@@ -193,11 +193,6 @@ function average(dvec::AbstractVector{UnivariateFinite{S,V,R,P}};
     return UnivariateFinite(sample_scitype(d1), d1.decoder, prob_given_ref)
 end
 
-
-function _pdf(d::UnivariateFinite{S,V,R,P}, ref) where {S,V,R,P}
-    return get(d.prob_given_ref, ref, zero(P))
-end
-
 """
     Distributions.pdf(d::UnivariateFinite, x)
 
@@ -226,11 +221,18 @@ One can also do weighted fits:
 
 See also `classes`, `support`.
 """
-Distributions.pdf(d::UnivariateFinite, cv::CategoricalValue) = _pdf(d, int(cv))
-
 function Distributions.pdf(
-    d::UnivariateFinite{S,V,R,P}, c::V) where {S,V,R,P}
+    d::UnivariateFinite{S,V,R,P},
+    cv::CategoricalValue,
+) where {S,V,R,P}
+    return get(d.prob_given_ref, int(cv), zero(P))
+end
+Distributions.pdf(d::UnivariateFinite{S,V}, c::V) where {S,V} = _pdf(d, c)
+
+# Avoid method ambiguity errors with Distributions >= 0.24
+Distributions.pdf(d::UnivariateFinite{S,V}, c::V) where {S,V<:Real} = _pdf(d, c)
 
+function _pdf(d::UnivariateFinite, c)
     _classes = classes(d)
     c in _classes || throw(DomainError("Value $c not in pool. "))
     pool = CategoricalArrays.pool(_classes)
@@ -239,8 +241,10 @@ function Distributions.pdf(
 end
 
 Distributions.logpdf(d::UnivariateFinite, cv::CategoricalValue) = log(pdf(d,cv))
+Distributions.logpdf(d::UnivariateFinite{S,V}, c::V) where {S,V} = log(pdf(d, c))
 
-Distributions.logpdf(d::UnivariateFinite{S,V,R,P}, c::V) where {S,V,R,P} = log(pdf(d,c))
+# Avoid method ambiguity errors with Distributions >= 0.24
+Distributions.logpdf(d::UnivariateFinite{S,V}, c::V) where {S,V<:Real} = log(pdf(d, c))
 
 function Distributions.mode(d::UnivariateFinite)
     dic = d.prob_given_ref

test/univariate_finite/methods.jl

@@ -222,6 +222,11 @@ end
     @test pdf(d, 'f') == 0
     @test pdf(d, f) == 0
     @test_throws DomainError pdf(d, 'j')
+
+    # regression test
+    # https://github.com/alan-turing-institute/MLJBase.jl/pull/432/files#r502299301
+    d = UnivariateFinite(classes(categorical(UInt32[0, 1])), [0.4, 0.6])
+    @test pdf(d, UInt32(1)) == 0.6
 end
 
 @testset "approx for UnivariateFinite" begin