Run JuliaFormatter and add CI (#27)

* Run JuliaFormatter and add CI

* Fix some URLs

Co-authored-by: willtebbutt <[email protected]>

Author: rossviljoen

.JuliaFormatter.toml

@@ -0,0 +1 @@
+style = "blue"

.github/workflows/Format.yml

@@ -0,0 +1,20 @@
+name: Format suggestions
+
+on:
+  pull_request:
+
+jobs:
+  format:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: julia-actions/setup-julia@latest
+        with:
+          version: 1
+      - run: |
+          julia  -e 'using Pkg; Pkg.add("JuliaFormatter")'
+          julia  -e 'using JuliaFormatter; format("."; verbose=true)'
+      - uses: reviewdog/action-suggester@v1
+        with:
+          tool_name: JuliaFormatter
+          fail_on_error: true

docs/make.jl

@@ -3,15 +3,11 @@ using Documenter, BayesianLinearRegressors
 makedocs(;
     modules=[BayesianLinearRegressors],
     format=Documenter.HTML(),
-    pages=[
-        "Home" => "index.md",
-    ],
-    repo="https://github.com/willtebbutt/BayesianLinearRegressors.jl/blob/{commit}{path}#L{line}",
+    pages=["Home" => "index.md"],
+    repo="https://github.com/JuliaGaussianProcesses/BayesianLinearRegressors.jl/blob/{commit}{path}#L{line}",
     sitename="BayesianLinearRegressors.jl",
     authors="Will Tebbutt <[email protected]>",
     assets=String[],
 )
 
-deploydocs(;
-    repo="github.com/willtebbutt/BayesianLinearRegressors.jl",
-)
+deploydocs(; repo="github.com/JuliaGaussianProcesses/BayesianLinearRegressors.jl")

examples/nn-blr.jl

@@ -15,10 +15,7 @@ using Statistics: mean, std
 # Create an MLP that you might have seen in the 90s.
 Dlat = 50;
 W1, b1 = randn(Dlat, 1), randn(Dlat);
-ϕ = Chain(
-    x->reshape(x, 1, :),
-    x->tanh.(W1 * x .+ b1),
-)
+ϕ = Chain(x -> reshape(x, 1, :), x -> tanh.(W1 * x .+ b1))
 
 # Initialise the standard deviation of the observation noise. We will learn this.
 logσ = [log(1)]
@@ -34,7 +31,7 @@ blr = BayesianLinearRegressor(zeros(Dlat), Matrix{Float64}(I, Dlat, Dlat))
 function nn_blr_training_loop(x, y, pars, Nitr, opt)
     tr_nlml = Vector{Float64}(undef, Nitr)
     p = ProgressMeter.Progress(Nitr)
-    for itr  in 1:Nitr
+    for itr in 1:Nitr
         nlml, back = Zygote.forward(Zygote.Params(pars)) do
             -logpdf(blr(ϕ(x), exp(2 * logσ[1])), y)
         end
@@ -44,7 +41,7 @@ function nn_blr_training_loop(x, y, pars, Nitr, opt)
             Flux.Optimise.update!(opt, par, g[par])
         end
         showvalues = [(:itr, itr), (:nlml, nlml), (:σ_ε, exp(logσ[1]))]
-        ProgressMeter.next!(p; showvalues = showvalues)
+        ProgressMeter.next!(p; showvalues=showvalues)
     end
     return tr_nlml
 end
@@ -71,12 +68,15 @@ plt = plot();
 
 # Plot the true function with aleatoric uncertainty due to observation noise.
 plot!(plt, xte, sin.(xte); linecolor="purple", linewidth=1.5, label="sin");
-plot!(plt, xte, sin.(xte) .+ 0.3; linecolor="purple", linewidth=1.5, label="", );
+plot!(plt, xte, sin.(xte) .+ 0.3; linecolor="purple", linewidth=1.5, label="");
 plot!(plt, xte, sin.(xte) .- 0.3; linecolor="purple", linewidth=1.5, label="")
 
 # Visualise the posterior marginal uncertainty via 3σ error bars.
 m_te, σ_te = mean.(ypr_te), std.(ypr_te)
-plot!(plt, xte, [m_te m_te];
+plot!(
+    plt,
+    xte,
+    [m_te m_te];
     label="",
     fillrange=[m_te .+ 3 .* σ_te, m_te .- 3 .* σ_te],
     linewidth=0,
@@ -85,22 +85,15 @@ plot!(plt, xte, [m_te m_te];
 );
 
 # Visualise the posterior distribution over the latent function via samples.
-plot!(plt, xte, y_samples_te[:, 1];
-    linecolor="blue",
-    linewidth=0.1,
-    label="Posterior Samples",
-);
-plot!(plt, xte, y_samples_te;
-    linecolor="blue",
-    linewidth=0.1,
-    label="",
+plot!(
+    plt, xte, y_samples_te[:, 1]; linecolor="blue", linewidth=0.1, label="Posterior Samples"
 );
+plot!(plt, xte, y_samples_te; linecolor="blue", linewidth=0.1, label="");
 
 # Plot the data.
 scatter!(plt, x, y; markercolor="red", label="y", markersize=0.1);
 display(plt);
 
-
 # Compute and display residuals
 ypr = marginals(blr′(ϕ(x), exp(2 * logσ[1])));
 ε = y .- mean.(ypr);

src/bayesian_linear_regression.jl

@@ -8,7 +8,7 @@ f(x) = dot(x, w)
 ```
 where `mw` and `Λw` are the mean and precision of `w`, respectively.
 """
-struct BayesianLinearRegressor{Tmw<:AbstractVector, TΛw<:AbstractMatrix} <: AbstractGP
+struct BayesianLinearRegressor{Tmw<:AbstractVector,TΛw<:AbstractMatrix} <: AbstractGP
     mw::Tmw
     Λw::TΛw
 end
@@ -35,7 +35,7 @@ AbstractGPs.mean_and_var(fx::FiniteBLR) = (mean(fx), var(fx))
 
 function AbstractGPs.rand(rng::AbstractRNG, fx::FiniteBLR, samples::Int)
     w = fx.f.mw .+ _cholesky(fx.f.Λw).U \ randn(rng, size(fx.x.X, 1), samples)
-    y = fx.x.X' * w .+ _cholesky(fx.Σy).U' * randn(rng, size(fx.x.X, 2), samples)
+    return fx.x.X' * w .+ _cholesky(fx.Σy).U' * randn(rng, size(fx.x.X, 2), samples)
 end
 
 function AbstractGPs.logpdf(fx::FiniteBLR, y::AbstractVector{<:Real})
@@ -84,7 +84,11 @@ end
 (s::BLRFunctionSample)(X::ColVecs) = X.X's.w
 (s::BLRFunctionSample)(X::RowVecs) = X.X * s.w
 
-Random.Sampler(::Type{<:AbstractRNG}, blr::BayesianLinearRegressor, ::Random.Repetition) = blr
+function Random.Sampler(
+    ::Type{<:AbstractRNG}, blr::BayesianLinearRegressor, ::Random.Repetition
+)
+    return blr
+end
 
 function Random.rand(rng::AbstractRNG, blr::BayesianLinearRegressor)
     w = blr.mw .+ _cholesky(blr.Λw).U \ randn(rng, size(blr.mw))
@@ -97,10 +101,12 @@ function Random.rand(rng::AbstractRNG, blr::BayesianLinearRegressor, dims::Dims)
     return reshape(bs, dims)
 end
 
-function Random.rand!(rng::AbstractRNG, A::AbstractArray{<:BLRFunctionSample}, blr::BayesianLinearRegressor)
+function Random.rand!(
+    rng::AbstractRNG, A::AbstractArray{<:BLRFunctionSample}, blr::BayesianLinearRegressor
+)
     ws = blr.mw .+ _cholesky(blr.Λw).U \ randn(rng, (only(size(blr.mw)), prod(size(A))))
     for i in LinearIndices(A)
-        @inbounds A[i] = BLRFunctionSample(ws[:,i])
+        @inbounds A[i] = BLRFunctionSample(ws[:, i])
     end
     return A
 end

test/bayesian_linear_regression.jl

@@ -13,7 +13,7 @@ end
         X, f, Σy = generate_toy_problem(rng, N, D)
 
         AbstractGPs.TestUtils.test_finitegp_primary_and_secondary_public_interface(
-            rng, f(X, Σy),
+            rng, f(X, Σy)
         )
     end
     @testset "rand" begin
@@ -24,8 +24,8 @@ end
         Y = rand(rng, f(X, Σy), samples)
         m_empirical = mean(Y; dims=2)
         Σ_empirical = (Y .- mean(Y; dims=2)) * (Y .- mean(Y; dims=2))' ./ samples
-        @test mean(f(X, Σy)) ≈ m_empirical atol=1e-3 rtol=1e-3
-        @test cov(f(X, Σy)) ≈ Σ_empirical atol=1e-3 rtol=1e-3
+        @test mean(f(X, Σy)) ≈ m_empirical atol = 1e-3 rtol = 1e-3
+        @test cov(f(X, Σy)) ≈ Σ_empirical atol = 1e-3 rtol = 1e-3
 
         @testset "Zygote (everything dense)" begin
             function rand_blr(X, A_Σy, mw, A_Λw)
@@ -45,10 +45,10 @@ end
 
             # Verify adjoints via finite differencing.
             fdm = central_fdm(5, 1)
-            @test dX ≈ first(j′vp(fdm, X->rand_blr(X, A_Σy, mw, A_Λw), z̄, X))
-            @test dA_Σy ≈ first(j′vp(fdm, A_Σy->rand_blr(X, A_Σy, mw, A_Λw), z̄, A_Σy))
-            @test dmw ≈ first(j′vp(fdm, mw->rand_blr(X, A_Σy, mw, A_Λw), z̄, mw))
-            @test dA_Λw ≈ first(j′vp(fdm, A_Λw->rand_blr(X, A_Σy, mw, A_Λw), z̄, A_Λw))
+            @test dX ≈ first(j′vp(fdm, X -> rand_blr(X, A_Σy, mw, A_Λw), z̄, X))
+            @test dA_Σy ≈ first(j′vp(fdm, A_Σy -> rand_blr(X, A_Σy, mw, A_Λw), z̄, A_Σy))
+            @test dmw ≈ first(j′vp(fdm, mw -> rand_blr(X, A_Σy, mw, A_Λw), z̄, mw))
+            @test dA_Λw ≈ first(j′vp(fdm, A_Λw -> rand_blr(X, A_Σy, mw, A_Λw), z̄, A_Λw))
         end
     end
     @testset "logpdf" begin
@@ -79,11 +79,13 @@ end
 
             # Check correctness via finite differencing.
             fdm = central_fdm(5, 1)
-            @test dX ≈ first(j′vp(fdm, X->logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, X))
-            @test dA_Σy ≈ first(j′vp(fdm, A_Σy->logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, A_Σy))
-            @test dy ≈ first(j′vp(fdm, y->logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, y))
-            @test dmw ≈ first(j′vp(fdm, mw->logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, mw))
-            @test dA_Λw ≈ first(j′vp(fdm, A_Λw->logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, A_Λw))
+            @test dX ≈ first(j′vp(fdm, X -> logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, X))
+            @test dA_Σy ≈
+                first(j′vp(fdm, A_Σy -> logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, A_Σy))
+            @test dy ≈ first(j′vp(fdm, y -> logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, y))
+            @test dmw ≈ first(j′vp(fdm, mw -> logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, mw))
+            @test dA_Λw ≈
+                first(j′vp(fdm, A_Λw -> logpdf_blr(X, A_Σy, y, mw, A_Λw), z̄, A_Λw))
         end
     end
     @testset "posterior" begin
@@ -105,14 +107,11 @@ end
             # Chop up the noise because we can't condition on noise that's correlated
             # between things.
             N1 = N - 3
-            Σ1, Σ2 = Σy[1:N1, 1:N1], Σy[N1+1:end, N1+1:end]
-            Σy′ = vcat(
-                hcat(Σ1, zeros(N1, N - N1)),
-                hcat(zeros(N - N1, N1), Σ2),
-            )
+            Σ1, Σ2 = Σy[1:N1, 1:N1], Σy[(N1 + 1):end, (N1 + 1):end]
+            Σy′ = vcat(hcat(Σ1, zeros(N1, N - N1)), hcat(zeros(N - N1, N1), Σ2))
 
-            X1, X2 = X[:, 1:N1], X[:, N1+1:end]
-            y1, y2 = y[1:N1], y[N1+1:end]
+            X1, X2 = X[:, 1:N1], X[:, (N1 + 1):end]
+            y1, y2 = y[1:N1], y[(N1 + 1):end]
 
             f′1 = posterior(f(X1, Σ1), y1)
             f′2 = posterior(f′1(X2, Σ2), y2)
@@ -135,7 +134,8 @@ end
         @test g(X) == g(Xr)
 
         # test the Random interface
-        @test rand(rng, Random.Sampler(rng, f, Val(Inf))) isa BayesianLinearRegressors.BLRFunctionSample
+        @test rand(rng, Random.Sampler(rng, f, Val(Inf))) isa
+            BayesianLinearRegressors.BLRFunctionSample
 
         samples1, samples2 = 10_000, 1000
         samples = samples1 * samples2
@@ -145,23 +145,21 @@ end
         # test statistical properties of the sampled functions
         let
             Y = reduce(hcat, map(h -> h(X), reshape(gs, :)))
-            m_empirical = mean(Y; dims = 2)
-            Σ_empirical = (Y .- mean(Y; dims = 2)) * (Y .- mean(Y; dims = 2))' ./ samples
+            m_empirical = mean(Y; dims=2)
+            Σ_empirical = (Y .- mean(Y; dims=2)) * (Y .- mean(Y; dims=2))' ./ samples
             @test mean(f(X, Σy)) ≈ m_empirical atol = 1e-3 rtol = 1e-3
             @test cov(f(X, Σy)) ≈ Σ_empirical + Σy atol = 1e-3 rtol = 1e-3
         end
 
         # test statistical properties of in-place rand
         let
             A = Array{BayesianLinearRegressors.BLRFunctionSample,2}(
-                undef,
-                samples1,
-                samples2,
+                undef, samples1, samples2
             )
             A = rand!(rng, A, f)
             Y = reduce(hcat, map(h -> h(X), reshape(gs, :)))
-            m_empirical = mean(Y; dims = 2)
-            Σ_empirical = (Y .- mean(Y; dims = 2)) * (Y .- mean(Y; dims = 2))' ./ samples
+            m_empirical = mean(Y; dims=2)
+            Σ_empirical = (Y .- mean(Y; dims=2)) * (Y .- mean(Y; dims=2))' ./ samples
             @test mean(f(X, Σy)) ≈ m_empirical atol = 1e-3 rtol = 1e-3
             @test cov(f(X, Σy)) ≈ Σ_empirical + Σy atol = 1e-3 rtol = 1e-3
         end
@@ -185,7 +183,7 @@ end
                 @test dX ≈ first(j′vp(fdm, X -> sample_function(X, mw, A_Λw), z̄, X))
                 @test dmw ≈ first(j′vp(fdm, mw -> sample_function(X, mw, A_Λw), z̄, mw))
                 @test dA_Λw ≈
-                      first(j′vp(fdm, A_Λw -> sample_function(X, mw, A_Λw), z̄, A_Λw))
+                    first(j′vp(fdm, A_Λw -> sample_function(X, mw, A_Λw), z̄, A_Λw))
             end
 
             function rand_funcs_single(X, mw, A_Λw)