final fixes

Project.toml

@@ -7,6 +7,7 @@ version = "0.4.0"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 ScopedValues = "7e506255-f358-4e82-b7e4-beb19740aa63"
+VectorInterface = "409d34a3-91d5-4945-b6ec-7529ddf182d8"
 
 [compat]
 Aqua = "0.8"
@@ -15,6 +16,7 @@ Printf = "1"
 Random = "1"
 ScopedValues = "1"
 Test = "1"
+VectorInterface = "0.5"
 julia = "1.8"
 
 [extras]

src/OptimKit.jl

@@ -3,6 +3,7 @@ module OptimKit
 using LinearAlgebra: LinearAlgebra
 using Printf
 using ScopedValues
+using VectorInterface
 using Base: @kwdef
 
 # Default values for the keyword arguments using ScopedValues
@@ -14,15 +15,34 @@ const GRADTOL = ScopedValue(1e-8)
 const MAXITER = ScopedValue(1_000_000)
 const VERBOSITY = ScopedValue(1)
 
-_retract(x, d, α) = (x + α * d, d)
-_inner(x, v1, v2) = v1 === v2 ? LinearAlgebra.norm(v1)^2 : LinearAlgebra.dot(v1, v2)
+# Default values for the manifold structure
+_retract(x, d, α) = (add(x, d, α), d)
+_inner(x, v1, v2) = v1 === v2 ? norm(v1)^2 : real(inner(v1, v2))
 _transport!(v, xold, d, α, xnew) = v
-_scale!(v, α) = LinearAlgebra.rmul!(v, α)
-_add!(vdst, vsrc, α) = LinearAlgebra.axpy!(α, vsrc, vdst)
+_scale!(v, α) = scale!!(v, α)
+_add!(vdst, vsrc, α) = add!!(vdst, vsrc, α)
 
 _precondition(x, g) = deepcopy(g)
 _finalize!(x, f, g, numiter) = x, f, g
 
+# Default structs for new convergence and termination keywords
+@kwdef struct DefaultHasConverged{T<:Real}
+    gradtol::T
+end
+
+function (d::DefaultHasConverged)(x, f, g, normgrad)
+    return normgrad <= d.gradtol
+end
+
+@kwdef struct DefaultShouldStop
+    maxiter::Int
+end
+
+function (d::DefaultShouldStop)(x, f, g, numfg, numiter, t)
+    return numiter >= d.maxiter
+end
+
+# Optimization
 abstract type OptimizationAlgorithm end
 
 const _xlast = Ref{Any}()
@@ -85,7 +105,6 @@ Also see [`GradientDescent`](@ref), [`ConjugateGradient`](@ref), [`LBFGS`](@ref)
 function optimize end
 
 include("linesearches.jl")
-include("terminate.jl")
 include("gd.jl")
 include("cg.jl")
 include("lbfgs.jl")

test/runtests.jl

@@ -49,17 +49,30 @@ function quadraticproblem(B, y)
     return fg
 end
 
+function quadratictupleproblem(B, y)
+    function fg(x)
+        x1, x2 = x
+        y1, y2 = y
+        g1 = B * (x1 - y1)
+        g2 = x2 - y2
+        f = dot(x1 - y1, g1) / 2 + (x2 - y2)^2 / 2
+        return f, (g1, g2)
+    end
+    return fg
+end
+
 algorithms = (GradientDescent, ConjugateGradient, LBFGS)
 
 @testset "Optimization Algorithm $algtype" for algtype in algorithms
     n = 10
     y = randn(n)
     A = randn(n, n)
-    fg = quadraticproblem(A' * A, y)
+    A = A' * A
+    fg = quadraticproblem(A, y)
     x₀ = randn(n)
     alg = algtype(; verbosity=2, gradtol=1e-12, maxiter=10_000_000)
     x, f, g, numfg, normgradhistory = optimize(fg, x₀, alg)
-    @test x ≈ y rtol = 10 * cond(A) * 1e-12
+    @test x ≈ y rtol = cond(A) * 1e-12
     @test f < 1e-12
 
     n = 1000
@@ -68,11 +81,12 @@ algorithms = (GradientDescent, ConjugateGradient, LBFGS)
     smax = maximum(S)
     A = U * Diagonal(1 .+ S ./ smax) * U'
     # well conditioned, all eigenvalues between 1 and 2
-    fg = quadraticproblem(A' * A, y)
-    x₀ = randn(n)
+    fg = quadratictupleproblem(A' * A, (y, 1.0))
+    x₀ = (randn(n), 2.0)
     alg = algtype(; verbosity=3, gradtol=1e-8)
     x, f, g, numfg, normgradhistory = optimize(fg, x₀, alg)
-    @test x ≈ y rtol = 1e-7
+    @test x[1] ≈ y rtol = 1e-7
+    @test x[2] ≈ 1 rtol = 1e-7
     @test f < 1e-12
 end