Add minimum, maximum, extrema for AbstractMvNormal and Product

docs/src/multivariate.md

@@ -71,6 +71,9 @@ invcov(::Distributions.AbstractMvNormal)
 logdetcov(::Distributions.AbstractMvNormal)
 sqmahal(::Distributions.AbstractMvNormal, ::AbstractArray)
 rand(::AbstractRNG, ::Distributions.AbstractMvNormal)
+minimum(::Distributions.AbstractMvNormal)
+maximum(::Distributions.AbstractMvNormal)
+extrema(::Distributions.AbstractMvNormal)
 ```
 
 ### MvLogNormal

src/common.jl

@@ -133,6 +133,12 @@ value_support(::Type{T}) where {T<:Distribution} = value_support(supertype(T))
 # to be decided: how to handle multivariate/matrixvariate distributions?
 Broadcast.broadcastable(d::UnivariateDistribution) = Ref(d)
 
+"""
+    extrema(d::Distribution)
+
+Return the minimum and maximum of the support of `d` as a 2-tuple.
+"""
+Base.extrema(d::Distribution) = minimum(d), maximum(d)
 
 ## TODO: the following types need to be improved
 abstract type SufficientStats end

src/multivariate/mvnormal.jl

@@ -80,6 +80,8 @@ abstract type AbstractMvNormal <: ContinuousMultivariateDistribution end
 insupport(d::AbstractMvNormal, x::AbstractVector) =
     length(d) == length(x) && all(isfinite, x)
 
+minimum(d::AbstractMvNormal) = fill(eltype(d)(-Inf), length(d))
+maximum(d::AbstractMvNormal) = fill(eltype(d)(Inf), length(d))
 mode(d::AbstractMvNormal) = mean(d)
 modes(d::AbstractMvNormal) = [mean(d)]
 

src/multivariate/product.jl

@@ -40,6 +40,8 @@ var(d::Product) = var.(d.v)
 cov(d::Product) = Diagonal(var(d))
 entropy(d::Product) = sum(entropy, d.v)
 insupport(d::Product, x::AbstractVector) = all(insupport.(d.v, x))
+minimum(d::Product) = map(minimum, d.v)
+maximum(d::Product) = map(maximum, d.v)
 
 """
     product_distribution(dists::AbstractVector{<:UnivariateDistribution})

src/univariates.jl

@@ -78,26 +78,19 @@ Get the degrees of freedom.
 dof(d::UnivariateDistribution)
 
 """
-    minimum(d::UnivariateDistribution)
+    minimum(d::Distribution)
 
 Return the minimum of the support of `d`.
 """
 minimum(d::UnivariateDistribution)
 
 """
-    maximum(d::UnivariateDistribution)
+    maximum(d::Distribution)
 
 Return the maximum of the support of `d`.
 """
 maximum(d::UnivariateDistribution)
 
-"""
-    extrema(d::UnivariateDistribution)
-
-Return the minimum and maximum of the support of `d` as a 2-tuple.
-"""
-extrema(d::UnivariateDistribution) = (minimum(d), maximum(d))
-
 """
     insupport(d::UnivariateDistribution, x::Any)
 

test/mvnormal.jl

@@ -29,6 +29,10 @@ function test_mvnormal(g::AbstractMvNormal, n_tsamples::Int=10^6,
     vs = diag(Σ)
     @test g == typeof(g)(params(g)...)
     @test g == deepcopy(g)
+    @test minimum(g) == fill(-Inf, d)
+    @test maximum(g) == fill(Inf, d)
+    @test extrema(g) == (minimum(g), maximum(g))
+    @test isless(extrema(g)...)
 
     # test sampling for AbstractMatrix (here, a SubArray):
     if ismissing(rng)

test/product.jl

@@ -29,7 +29,7 @@ end
     N = 11
     # Construct independent distributions and `Product` distribution from these.
     ubound = rand(N)
-    ds = Uniform.(0.0, ubound)
+    ds = Uniform.(-ubound, ubound)
     x = rand.(ds)
     d_product = product_distribution(ds)
     @test d_product isa Product
@@ -43,6 +43,10 @@ end
     @test entropy(d_product) == sum(entropy.(ds))
     @test insupport(d_product, ubound) == true
     @test insupport(d_product, ubound .+ 1) == false
+    @test minimum(d_product) == -ubound
+    @test maximum(d_product) == ubound
+    @test extrema(d_product) == (-ubound, ubound)
+    @test isless(extrema(d_product)...)
 
     y = rand(d_product)
     @test y isa typeof(x)