Update maximize.jl (#1119)

* Update maximize.jl

* Update maximize.jl

* Update maximize.jl

Author: pkofod

src/maximize.jl

@@ -38,7 +38,7 @@ end
 function maximize(f, g, h, x0::AbstractArray, method::AbstractOptimizer, options = Optim.Options(); kwargs...)
     fmax = x->-f(x)
     gmax = (G,x)->(g(G,x); G.=-G)
-    hmax = (H,x)->(h(G,x); H.=-H)
+    hmax = (H,x)->(h(H,x); H.=-H)
     MaximizationWrapper(optimize(fmax, gmax, hmax, x0, method, options; kwargs...))
 end
 

test/general/maximize.jl

@@ -1,26 +1,79 @@
 @testset "maximization wrapper" begin
     @testset "univariate" begin
-        resmax = maximize(x->x^3, -1, 9)
-        resmin = optimize(x->-x^3, -1, 9)
+        resmax = maximize(x -> x^3, -1, 9)
+        resmin = optimize(x -> -x^3, -1, 9)
         @test Optim.maximum(resmax) == -Optim.minimum(resmin)
         @test resmax.res.minimum == resmin.minimum
         for meth in (Brent(), GoldenSection())
-            resmax = maximize(x->x^3, -1, 9, meth)
-            resmin = optimize(x->-x^3, -1, 9, meth)
+            resmax = maximize(x -> x^3, -1, 9, meth)
+            resmin = optimize(x -> -x^3, -1, 9, meth)
             @test Optim.maximum(resmax) == -Optim.minimum(resmin)
             @test resmax.res.minimum == resmin.minimum
         end
     end
     @testset "multivariate" begin
-        resmax = maximize(x->x[1]^3+x[2]^2, [3.0, 0.0])
-        resmin = optimize(x->-x[1]^3-x[2]^2, [3.0, 0.0])
+        resmax = maximize(x -> x[1]^3 + x[2]^2, [3.0, 0.0])
+        resmin = optimize(x -> -x[1]^3 - x[2]^2, [3.0, 0.0])
         @test Optim.maximum(resmax) == -Optim.minimum(resmin)
         @test resmax.res.minimum == resmin.minimum
-        for meth in (NelderMead(), BFGS(), LBFGS(), GradientDescent(), Newton(), NewtonTrustRegion(), SimulatedAnnealing())
-            resmax = maximize(x->x[1]^3+x[2]^2, [3.0, 0.0])
-            resmin = optimize(x->-x[1]^3-x[2]^2, [3.0, 0.0])
+        for meth in (
+            NelderMead(),
+            BFGS(),
+            LBFGS(),
+            GradientDescent(),
+            Newton(),
+            NewtonTrustRegion(),
+            SimulatedAnnealing(),
+        )
+            resmax = maximize(x -> x[1]^3 + x[2]^2, [3.0, 0.0])
+            resmin = optimize(x -> -x[1]^3 - x[2]^2, [3.0, 0.0])
             @test Optim.maximum(resmax) == -Optim.minimum(resmin)
             @test resmax.res.minimum == resmin.minimum
         end
     end
+
+    prob = MVP.UnconstrainedProblems.examples["Powell"]
+    f = objective(prob)
+    g! = gradient(prob)
+    h! = hessian(prob)
+    fmax(x) = -f(x)
+    gmax = (G, x) -> (g!(G, x); G .= -G)
+    hmax = (H, x) -> (h!(H, x); H .= -H)
+
+    resmax_f = maximize(fmax, prob.initial_x)
+    resmin_f = optimize(f, prob.initial_x)
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_f.res, prop) == getproperty(resmin_f, prop)
+    end
+
+    resmax_f_nm = maximize(fmax, prob.initial_x, NelderMead())
+    resmin_f_nm = optimize(f, prob.initial_x, NelderMead())
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_f_nm.res, prop) == getproperty(resmin_f_nm, prop)
+    end
+
+    resmax_f_bfgs = maximize(fmax, prob.initial_x, BFGS())
+    resmin_f_bfgs = optimize(f, prob.initial_x, BFGS())
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_f_bfgs.res, prop) == getproperty(resmin_f_bfgs, prop)
+    end
+
+    resmax_f_newton = maximize(fmax, prob.initial_x, Newton())
+    resmin_f_newton = optimize(f, prob.initial_x, Newton())
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_f_newton.res, prop) == getproperty(resmin_f_newton, prop)
+    end
+
+    resmax_fg = maximize(fmax, gmax, prob.initial_x, BFGS())
+    resmin_fg = optimize(f, g!, prob.initial_x, BFGS())
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_fg.res, prop) == getproperty(resmin_fg, prop)
+    end
+
+    resmax_fgh = maximize(fmax, gmax, hmax, prob.initial_x, Newton())
+    resmin_fgh = optimize(f, g!, h!, prob.initial_x, Newton())
+    for prop in (:iterations, :ls_success, :minimizer, :minimum)
+        @test getproperty(resmax_fgh.res, prop) == getproperty(resmin_fgh, prop)
+    end
+
 end