Add unified interface test utils

test/likelihoods/bernoulli.jl

@@ -1,13 +1,4 @@
 @testset "BernoulliLikelihood" begin
-    rng = MersenneTwister(123)
-    gp = GP(SqExponentialKernel())
-    x = rand(rng, 10)
-    y = rand(rng, 10)
     lik = BernoulliLikelihood()
-    lgp = LatentGP(gp, lik, 1e-5)
-    lfgp = lgp(x)
-
-    @test typeof(lik(rand(rng, lfgp.fx))) <: Distribution
-    @test length(rand(rng, lik(rand(rng, lfgp.fx)))) == 10
-    @test Functors.functor(lik)[1] == ()
+    test_interface(lik, SqExponentialKernel(), rand(10))
 end

test/likelihoods/categorical.jl

@@ -1,22 +1,8 @@
 @testset "CategoricalLikelihood" begin
-    rng = MersenneTwister(123)
-    gp = GP(IndependentMOKernel(SqExponentialKernel()))
+    lik = CategoricalLikelihood()
     IN_DIM = 3
-    OUT_DIM = 4
+    OUT_DIM = 2 # one for the mean the other for the log-standard deviation
     N = 10
-    x = [rand(rng, IN_DIM) for _=1:N]
-    X = MOInput(x, OUT_DIM)
-    lik = CategoricalLikelihood()
-    lgp = LatentGP(gp, lik, 1e-5)
-    lfgp = lgp(X)
-
-    Y = rand(rng, lfgp.fx)
-
-    y = [Y[[i + j*N for j in 0:(OUT_DIM - 1)]] for i in 1:N]
-    # Replace with mo_inverse_transform once it is merged
-
-    @test length(lik(rand(3)).p) == 4
-    @test lik(y) isa Distribution
-    @test length(rand(rng, lik(y))) == 10
-    @test Functors.functor(lik)[1] == ()
+    X = MOInput([rand(IN_DIM) for _ in 1:N], OUT_DIM)
+    test_interface(lik, IndependentMOKernel(SqExponentialKernel()), X)
 end

test/likelihoods/gaussian.jl

@@ -1,35 +1,13 @@
 @testset "GaussianLikelihood" begin
-    rng = MersenneTwister(123)
-    gp = GP(SqExponentialKernel())
-    x = rand(rng, 10)
-    y = rand(rng, 10)
     lik = GaussianLikelihood(1e-5)
-    lgp = LatentGP(gp, lik, 1e-5)
-    lfgp = lgp(x)
-
-    @test lik(rand(rng, lfgp.fx)) isa Distribution
-    @test length(rand(rng, lik(rand(rng, lfgp.fx)))) == 10
-    @test keys(Functors.functor(lik)[1]) == (:σ²,)
+    test_interface(lik, SqExponentialKernel(), rand(10); functor_args=(:σ²,))
 end
 
 @testset "HeteroscedasticGaussianLikelihood" begin
-    rng = MersenneTwister(123)
-    gp = GP(IndependentMOKernel(SqExponentialKernel()))
+    lik = HeteroscedasticGaussianLikelihood()
     IN_DIM = 3
     OUT_DIM = 2 # one for the mean the other for the log-standard deviation
     N = 10
-    x = [rand(rng, IN_DIM) for _ in 1:N]
-    X = MOInput(x, OUT_DIM)
-    lik = HeteroscedasticGaussianLikelihood()
-    lgp = LatentGP(gp, lik, 1e-5)
-    lfgp = lgp(X)
-
-    Y = rand(rng, lfgp.fx)
-
-    y = [Y[[i + j*N for j in 0:(OUT_DIM - 1)]] for i in 1:N]
-    # Replace with mo_inverse_transform once it is merged
-
-    @test lik(y) isa Distribution
-    @test length(rand(rng, lik(y))) == 10
-    @test Functors.functor(lik)[1] == ()
+    X = MOInput([rand(IN_DIM) for _ in 1:N], OUT_DIM)
+    test_interface(lik, IndependentMOKernel(SqExponentialKernel()), X)
 end

test/likelihoods/poisson.jl

@@ -1,13 +1,4 @@
 @testset "PoissonLikelihood" begin
-    rng = MersenneTwister(123)
-    gp = GP(SqExponentialKernel())
-    x = rand(rng, 10)
-    y = rand(rng, 10)
     lik = PoissonLikelihood()
-    lgp = LatentGP(gp, lik, 1e-5)
-    lfgp = lgp(x)
-
-    @test lik(rand(rng, lfgp.fx)) isa Distribution
-    @test length(rand(rng, lik(rand(rng, lfgp.fx)))) == 10
-    @test Functors.functor(lik)[1] == ()
+    test_interface(lik, SqExponentialKernel(), rand(10))
 end

test/runtests.jl

@@ -7,6 +7,8 @@ using Distributions
 
 @testset "GPLikelihoods.jl" begin
 
+    include("test_utils.jl")
+
     @testset "likelihoods" begin
         include("likelihoods/bernoulli.jl")
         include("likelihoods/categorical.jl")

test/test_utils.jl

@@ -0,0 +1,60 @@
+function test_interface(
+    rng::AbstractRNG,
+    lik,
+    k::KernelFunctions.Kernel,
+    x::AbstractVector;
+    functor_args=(),
+)
+    gp = GP(k)
+    lgp = LatentGP(gp, lik, 1e-5)
+    lfgp = lgp(x)
+
+    # Check if likelihood produces a distribution
+    @test lik(rand(rng, lfgp.fx)) isa Distribution
+
+    N = length(x)
+    y = rand(rng, lfgp.fx)
+
+    if x isa MOInput
+        # TODO: replace with mo_inverse_transform
+        N = length(x.x)
+        y = [y[[i + j*N for j in 0:(x.out_dim - 1)]] for i in 1:N]
+    end
+
+    # Check if the likelihood samples are of correct length
+    @test length(rand(rng, lik(y))) == N
+
+    # Check if functor works properly
+    if functor_args == ()
+        @test Functors.functor(lik)[1] == functor_args
+    else
+        @test keys(Functors.functor(lik)[1]) == functor_args
+    end
+end
+
+"""
+    test_interface(
+        lik,
+        k::Kernel,
+        x::AbstractVector;
+        functor_args=(),
+    )
+
+This function provides unified method to check the interface of the various likelihoods 
+defined. It checks if the likelihood produces a distribution, length of likelihood 
+samples is correct and if the functor works as intended.  
+"""
+function test_interface(
+    lik,
+    k::KernelFunctions.Kernel,
+    x::AbstractVector;
+    kwargs...
+)
+    test_interface(
+        Random.GLOBAL_RNG,
+        lik,
+        k,
+        x;
+        kwargs...
+    )
+end