Updates for Julia 0.6

.travis.yml

@@ -3,8 +3,8 @@ os:
     - osx
     - linux
 julia:
-    - 0.4
     - 0.5
+    - 0.6
     - nightly
 notifications:
     email: false

REQUIRE

@@ -1,3 +1,3 @@
-julia 0.4
-Compat 0.8.4
+julia 0.5
+Compat 0.17.0
 StatsBase 0.7.0

src/MultivariateStats.jl

@@ -4,7 +4,7 @@ module MultivariateStats
     using Compat
     using StatsBase
 
-    import Base: length, size, show, dump, sumabs2
+    import Base: length, size, show, dump
     import Base.LinAlg: Cholesky
     import StatsBase: fit, predict
     using  Compat: view

src/cmds.jl

@@ -37,7 +37,7 @@ function dmat2gram!{GT}(G::AbstractMatrix{GT}, D::AbstractMatrix)
     u = zeros(GT, n)
     s = 0.0
     for j = 1:n
-        s += (u[j] = Base.sumabs2(view(D,:,j)) / n)
+        s += (u[j] = sum(abs2, view(D,:,j)) / n)
     end
     s /= n
 
@@ -98,8 +98,8 @@ function classical_mds{T<:Real}(D::AbstractMatrix{T}, p::Int;
 
     #Check if the last considered eigenvalue is degenerate
     if m>0
-        nevalsmore = sum(abs(E[:values][ord[m+1:end]] .- v[m]^2) .< n*eps())
-        nevals = sum(abs(E[:values] .- v[m]^2) .< n*eps())
+        nevalsmore = sum(abs.(E[:values][ord[m+1:end]] .- v[m]^2) .< n*eps())
+        nevals = sum(abs.(E[:values] .- v[m]^2) .< n*eps())
         if nevalsmore > 1
             dowarn && warn("The last eigenpair is degenerate with $(nevals-1) others; $nevalsmore were ignored. Answer is not unique")
         end

src/ica.jl

@@ -24,7 +24,7 @@ transform(M::ICA, x::AbstractVecOrMat) = At_mul_B(M.W, centralize(x, M.mean))
 #
 # It returns a function value v, and derivative d
 #
-abstract ICAGDeriv
+@compat abstract type ICAGDeriv end
 
 immutable Tanh <: ICAGDeriv
     a::Float64
@@ -76,15 +76,15 @@ function fastica!(W::DenseMatrix{Float64},      # initialized component matrix,
     end
 
     # pre-allocated storage
-    Wp = Array(Float64, m, k)    # to store previous version of W
-    U  = Array(Float64, n, k)    # to store w'x & g(w'x)
-    Y  = Array(Float64, m, k)    # to store E{x g(w'x)} for components
-    E1 = Array(Float64, k)       # store E{g'(w'x)} for components
+    Wp = Matrix{Float64}(m, k)    # to store previous version of W
+    U  = Matrix{Float64}(n, k)    # to store w'x & g(w'x)
+    Y  = Matrix{Float64}(m, k)    # to store E{x g(w'x)} for components
+    E1 = Vector{Float64}(k)       # store E{g'(w'x)} for components
 
     # normalize each column
     for j = 1:k
         w = view(W,:,j)
-        scale!(w, 1.0 / sqrt(sumabs2(w)))
+        scale!(w, 1.0 / sqrt(sum(abs2, w)))
     end
 
     # main loop
@@ -178,14 +178,14 @@ function fit(::Type{ICA}, X::DenseMatrix{Float64},          # sample matrix, siz
         Efac = eigfact(C)
         ord = sortperm(Efac.values; rev=true)
         (v, P) = extract_kv(Efac, ord, k)
-        W0 = scale!(P, 1.0 ./ sqrt(v))
+        W0 = scale!(P, 1.0 ./ sqrt.(v))
         # println(W0' * C * W0)
         Z = W0'Z
     end
 
     # initialize
     W = (isempty(winit) ? randn(size(Z,1), k) : copy(winit))::Matrix{Float64}
-    
+
     # invoke core algorithm
     fastica!(W, Z, fun, maxiter, tol, verbose)
 

src/lda.jl

@@ -2,7 +2,7 @@
 
 #### Type to represent a linear discriminant functional
 
-abstract Discriminant{T}
+@compat abstract type Discriminant{T} end
 
 immutable LinearDiscriminant{T<:AbstractFloat} <: Discriminant{T}
     w::Vector{T}
@@ -105,7 +105,7 @@ function multiclass_lda_stats{T<:AbstractFloat}(nc::Int, X::DenseMatrix{T}, y::A
 
     # compute between-class scattering
     mean = cmeans * (cweights ./ n)
-    U = scale!(cmeans .- mean, sqrt(cweights))
+    U = scale!(cmeans .- mean, sqrt.(cweights))
     Sb = A_mul_Bt(U, U)
 
     return MulticlassLDAStats(Vector{T}(cweights), mean, cmeans, Sw, Sb)
@@ -229,7 +229,7 @@ function fit{T,F<:SubspaceLDA}(::Type{F}, X::AbstractMatrix{T}, label::AbstractV
     # Note Sb = Hb*Hb', Sw = Hw*Hw'
     cmeans, cweights, Hw = center(X, label, nc)
     dmeans = cmeans .- (normalize ? mean(cmeans, 2) : cmeans * (cweights / n))
-    Hb = normalize ? dmeans : dmeans * Diagonal(sqrt(cweights))
+    Hb = normalize ? dmeans : dmeans * Diagonal(sqrt.(cweights))
     if normalize
         Hw /= sqrt(n)
     end
@@ -260,7 +260,7 @@ function lda_gsvd{T}(Hb::AbstractMatrix{T}, Hw::AbstractMatrix{T}, cweights::Abs
     # Normalize
     Gw = G' * Hw
     nrm = Gw * Gw'
-    G = G ./ reshape(sqrt(diag(nrm)), 1, ncnz-1)
+    G = G ./ reshape(sqrt.(diag(nrm)), 1, ncnz-1)
     # Also get the eigenvalues
     Gw = G' * Hw
     Gb = G' * Hb
@@ -288,7 +288,7 @@ function center{T}(X::AbstractMatrix{T}, label::AbstractVector{Int}, nc=maximum(
         end
     end
     # Compute differences from the means
-    dX = Array(T, d, n)
+    dX = Matrix{T}(d, n)
     for j = 1:n
         k = label[j]
         for i = 1:d

src/pca.jl

@@ -148,11 +148,7 @@ function fit{T<:AbstractFloat}(::Type{PCA}, X::DenseMatrix{T};
 
     # delegate to core
     if method == :cov
-        if VERSION < v"0.5.0-dev+660"
-            C = cov(X; vardim=2, mean=isempty(mv) ? 0 : mv)::Matrix{T}
-        else
-            C = Base.covm(X, isempty(mv) ? 0 : mv, 2)
-        end
+        C = Base.covm(X, isempty(mv) ? 0 : mv, 2)
         M = pcacov(C, mv; maxoutdim=maxoutdim, pratio=pratio)
     elseif method == :svd
         Z = centralize(X, mv)

src/ppca.jl

@@ -180,11 +180,7 @@ function fit{T<:AbstractFloat}(::Type{PPCA}, X::DenseMatrix{T};
         Z = centralize(X, mv)
         M = ppcaml(Z, mv, maxoutdim=maxoutdim, tol=tol)
     elseif method == :em || method == :bayes
-        S = if VERSION < v"0.5.0-dev+660"
-            cov(X; vardim=2, mean=isempty(mv) ? 0 : mv)::Matrix{T}
-        else
-            Base.covm(X, isempty(mv) ? 0 : mv, 2)
-        end
+        S = Base.covm(X, isempty(mv) ? 0 : mv, 2)
         if method == :em
             M = ppcaem(S, mv, n, maxoutdim=maxoutdim, tol=tol, tot=tot)
         elseif method == :bayes

src/whiten.jl

@@ -6,11 +6,7 @@
 #
 function cov_whitening{T<:AbstractFloat}(C::Cholesky{T})
     cf = C[:UL]
-    if VERSION < v"0.4.0-"
-        full(istriu(cf) ? inv(Triangular(C.UL, :U)) : inv(Triangular(C.UL, :L)'))::Matrix{T}
-    else
-        full(inv(istriu(cf) ? cf : cf'))::Matrix{T}
-    end
+    full(inv(istriu(cf) ? cf : cf'))::Matrix{T}
 end
 
 cov_whitening!{T<:AbstractFloat}(C::DenseMatrix{T}) = cov_whitening(cholfact!(C, :U))
@@ -28,15 +24,15 @@ cov_whitening{T<:AbstractFloat}(C::DenseMatrix{T}, regcoef::Real) =
 immutable Whitening{T<:AbstractFloat}
     mean::Vector{T}
     W::Matrix{T}
-    function Whitening(mean::Vector{T}, W::Matrix{T})
+    function (::Type{Whitening{T}}){T}(mean::Vector{T}, W::Matrix{T})
         d, d2 = size(W)
         d == d2 || error("W must be a square matrix.")
         isempty(mean) || length(mean) == d ||
         throw(DimensionMismatch("Sizes of mean and W are inconsistent."))
-        return new(mean, W)
+        return new{T}(mean, W)
     end
 end
-Whitening{T<:AbstractFloat}(mean::Vector{T}, W::Matrix{T}) = Whitening{T}(mean, W)
+(::Type{Whitening}){T<:AbstractFloat}(mean::Vector{T}, W::Matrix{T}) = Whitening{T}(mean, W)
 
 indim(f::Whitening) = size(f.W, 1)
 outdim(f::Whitening) = size(f.W, 2)

test/cca.jl

@@ -27,8 +27,8 @@ M = CCA(Float64[], Float64[], Px, Py, [0.8, 0.6, 0.4])
 @test yprojection(M) == Py
 @test correlations(M) == [0.8, 0.6, 0.4]
 
-@test_approx_eq xtransform(M, X) Px'X
-@test_approx_eq ytransform(M, Y) Py'Y
+@test xtransform(M, X) ≈ Px'X
+@test ytransform(M, Y) ≈ Py'Y
 
 ## CCA with nonzero means
 
@@ -45,8 +45,8 @@ M = CCA(ux, uy, Px, Py, [0.8, 0.6, 0.4])
 @test yprojection(M) == Py
 @test correlations(M) == [0.8, 0.6, 0.4]
 
-@test_approx_eq xtransform(M, X) Px' * (X .- ux)
-@test_approx_eq ytransform(M, Y) Py' * (Y .- uy)
+@test xtransform(M, X) ≈ Px' * (X .- ux)
+@test ytransform(M, Y) ≈ Py' * (Y .- uy)
 
 
 ## prepare data
@@ -62,15 +62,9 @@ ym = vec(mean(Y, 2))
 Zx = X .- xm
 Zy = Y .- ym
 
-if VERSION < v"0.5.0-dev+660"
-    Cxx = cov(X; vardim=2)
-    Cyy = cov(Y; vardim=2)
-    Cxy = cov(X, Y; vardim=2)
-else
-    Cxx = cov(X, 2)
-    Cyy = cov(Y, 2)
-    Cxy = cov(X, Y, 2)
-end
+Cxx = cov(X, 2)
+Cyy = cov(Y, 2)
+Cxy = cov(X, Y, 2)
 Cyx = Cxy'
 
 ## ccacov
@@ -87,12 +81,12 @@ rho = correlations(M)
 @test ymean(M) == ym
 @test issorted(rho; rev=true)
 
-@test_approx_eq Px' * Cxx * Px eye(p)
-@test_approx_eq Py' * Cyy * Py eye(p)
-@test_approx_eq Cxy * (Cyy \ Cyx) * Px  Cxx * Px * Diagonal(rho.^2)
-@test_approx_eq Cyx * (Cxx \ Cxy) * Py  Cyy * Py * Diagonal(rho.^2)
-@test_approx_eq Px qnormalize!(Cxx \ (Cxy * Py), Cxx)
-@test_approx_eq Py qnormalize!(Cyy \ (Cyx * Px), Cyy)
+@test Px' * Cxx * Px ≈ eye(p)
+@test Py' * Cyy * Py ≈ eye(p)
+@test Cxy * (Cyy \ Cyx) * Px ≈ Cxx * Px * Diagonal(rho.^2)
+@test Cyx * (Cxx \ Cxy) * Py ≈ Cyy * Py * Diagonal(rho.^2)
+@test Px ≈ qnormalize!(Cxx \ (Cxy * Py), Cxx)
+@test Py ≈ qnormalize!(Cyy \ (Cyx * Px), Cyy)
 
 # Y ~ X
 M = fit(CCA, Y, X; method=:cov, outdim=p)
@@ -106,12 +100,12 @@ rho = correlations(M)
 @test ymean(M) == xm
 @test issorted(rho; rev=true)
 
-@test_approx_eq Px' * Cxx * Px eye(p)
-@test_approx_eq Py' * Cyy * Py eye(p)
-@test_approx_eq Cxy * (Cyy \ Cyx) * Px  Cxx * Px * Diagonal(rho.^2)
-@test_approx_eq Cyx * (Cxx \ Cxy) * Py  Cyy * Py * Diagonal(rho.^2)
-@test_approx_eq Px qnormalize!(Cxx \ (Cxy * Py), Cxx)
-@test_approx_eq Py qnormalize!(Cyy \ (Cyx * Px), Cyy)
+@test Px' * Cxx * Px ≈ eye(p)
+@test Py' * Cyy * Py ≈ eye(p)
+@test Cxy * (Cyy \ Cyx) * Px ≈ Cxx * Px * Diagonal(rho.^2)
+@test Cyx * (Cxx \ Cxy) * Py ≈ Cyy * Py * Diagonal(rho.^2)
+@test Px ≈ qnormalize!(Cxx \ (Cxy * Py), Cxx)
+@test Py ≈ qnormalize!(Cyy \ (Cyx * Px), Cyy)
 
 
 ## ccasvd
@@ -128,9 +122,9 @@ rho = correlations(M)
 @test ymean(M) == ym
 @test issorted(rho; rev=true)
 
-@test_approx_eq Px' * Cxx * Px eye(p)
-@test_approx_eq Py' * Cyy * Py eye(p)
-@test_approx_eq Cxy * (Cyy \ Cyx) * Px  Cxx * Px * Diagonal(rho.^2)
-@test_approx_eq Cyx * (Cxx \ Cxy) * Py  Cyy * Py * Diagonal(rho.^2)
-@test_approx_eq Px qnormalize!(Cxx \ (Cxy * Py), Cxx)
-@test_approx_eq Py qnormalize!(Cyy \ (Cyx * Px), Cyy)
+@test Px' * Cxx * Px ≈ eye(p)
+@test Py' * Cyy * Py ≈ eye(p)
+@test Cxy * (Cyy \ Cyx) * Px ≈ Cxx * Px * Diagonal(rho.^2)
+@test Cyx * (Cxx \ Cxy) * Py ≈ Cyy * Py * Diagonal(rho.^2)
+@test Px ≈ qnormalize!(Cxx \ (Cxy * Py), Cxx)
+@test Py ≈ qnormalize!(Cyy \ (Cyx * Px), Cyy)

test/cmds.jl

@@ -5,10 +5,10 @@ using Compat
 ## testing data
 
 function pwdists(X)
-	S = Base.sumabs2(X, 1)
-	D2 = S .+ S' - 2 * (X'X)
-	D2[diagind(D2)] = 0.0
-	return sqrt(D2)
+    S = sum(abs2, X, 1)
+    D2 = S .+ S' - 2 * (X'X)
+    D2[diagind(D2)] = 0.0
+    return sqrt.(D2)
 end
 
 X0 = randn(3, 5)
@@ -22,16 +22,16 @@ D0 = pwdists(X0)
 
 D = gram2dmat(G0)
 @test issymmetric(D)
-@test_approx_eq D D0
+@test D ≈ D0
 
 G = dmat2gram(D0)
 @test issymmetric(G)
-@test_approx_eq gram2dmat(G) D0
+@test gram2dmat(G) ≈ D0
 
 ## classical MDS
 
 X = classical_mds(D0, 3)
-@test_approx_eq pwdists(X) D0
+@test pwdists(X) ≈ D0
 
 #Test MDS embeddings in dimensions >= number of points
 @test classical_mds([0 1; 1 0], 2, dowarn=false) == [-0.5 0.5; 0 0]
@@ -52,12 +52,7 @@ let
  0.38095238095238093 0.4 0.21052631578947367 0.5454545454545454 0.3333333333333333 0.3181818181818182 0.23529411764705882 0.5555555555555556 0.4 1.0]
     X =  [-0.27529104101488666 0.006134513718202863 0.33298809606740326 0.2608994458893664 -0.46185275796909575 -0.23734315039370618 0.29972782027671513 0.03827901455843394 -0.04096713097883363 0.07742518984640051
  -0.08177061420820278 -0.0044504235228030225 -0.3271919093638943 0.28206254638779243 -0.0954706915166714 -0.07137742126520012 -0.30754933764853587 0.18582658369448027 -0.03715307349750036 0.45707434094053534]
-    if VERSION < v"0.4.0"
-        #Don't have correct eigenvector testing; run a cruder test
-        @test_approx_eq abs(classical_mds(D, 2)) abs(X)
-    else
-        Test.test_approx_eq_modphase(classical_mds(D, 2)', X')
-    end
+    Test.test_approx_eq_modphase(classical_mds(D, 2)', X')
 end
 
 #10 - test degenerate problem