Merge pull request #508 from sathvikbhagavan/sb/fix_rtea

test: revive RTEA optimization tests

Author: ChrisRackauckas

lib/SurrogatesSVM/src/SurrogatesSVM.jl

@@ -33,7 +33,7 @@ function SVMSurrogate(x, y, lb, ub)
         end
     else
         for j in eachindex(x)
-            X[j, :] = x[j]
+            X[j, :] .= x[j]
         end
     end
     model = LIBSVM.fit!(SVC(), X, y)
@@ -46,7 +46,7 @@ end
 
 function (svmsurr::SVMSurrogate)(val)
     n = length(val)
-    return LIBSVM.predict(svmsurr.model, reshape(val, 1, n))[1]
+    return LIBSVM.predict(svmsurr.model, reshape(collect(val), 1, n))[1]
 end
 
 """
@@ -65,7 +65,8 @@ function SurrogatesBase.update!(svmsurr::SVMSurrogate, x_new, y_new)
         svmsurr.model = LIBSVM.fit!(
             SVC(), reshape(svmsurr.x, length(svmsurr.x), 1), svmsurr.y)
     else
-        svmsurr.model = LIBSVM.fit!(SVC(), transpose(reduce(hcat, svmsurr.x)), svmsurr.y)
+        svmsurr.model = LIBSVM.fit!(
+            SVC(), transpose(reduce(hcat, collect.(svmsurr.x))), svmsurr.y)
     end
 end
 

lib/SurrogatesSVM/test/runtests.jl

@@ -25,16 +25,17 @@ using SafeTestsets
         obj_N = x -> x[1]^2 * x[2]
         lb = [0.0, 0.0]
         ub = [10.0, 10.0]
-        x = collect.(sample(100, lb, ub, RandomSample()))
+        x = sample(100, lb, ub, RandomSample())
         y = obj_N.(x)
-        svm = LIBSVM.fit!(SVC(), transpose(reduce(hcat, x)), y)
+        svm = LIBSVM.fit!(SVC(), transpose(reduce(hcat, collect.(x))), y)
         my_svm_ND = SVMSurrogate(x, y, lb, ub)
         x_test = [5.0, 1.2]
         val = my_svm_ND(x_test)
         @test LIBSVM.predict(svm, reshape(x_test, 1, 2))[1] == val
-        update!(my_svm_ND, [[1.0, 1.0]], [1.0])
-        update!(my_svm_ND, [[1.2, 1.2], [1.5, 1.5]], [1.728, 3.375])
-        svm = LIBSVM.fit!(SVC(), transpose(reduce(hcat, my_svm_ND.x)), my_svm_ND.y)
+        update!(my_svm_ND, [(1.0, 1.0)], [1.0])
+        update!(my_svm_ND, [(1.2, 1.2), (1.5, 1.5)], [1.728, 3.375])
+        svm = LIBSVM.fit!(
+            SVC(), transpose(reduce(hcat, collect.(my_svm_ND.x))), my_svm_ND.y)
         x_test = [1.0, 1.0]
         val = my_svm_ND(x_test)
         @test LIBSVM.predict(svm, reshape(x_test, 1, 2))[1] == val

src/VariableFidelity.jl

@@ -145,7 +145,7 @@ function (varfid::VariableFidelitySurrogate)(val)
 end
 
 """
-add_point!(varfid::VariableFidelitySurrogate,x_new,y_new)
+update!(varfid::VariableFidelitySurrogate,x_new,y_new)
 
 I expect to add low fidelity data to the surrogate.
 """

test/optimization.jl

@@ -1,6 +1,7 @@
 using Surrogates
 using LinearAlgebra
 using QuasiMonteCarlo
+using SurrogatesSVM
 #######SRBF############
 ##### 1D #####
 
@@ -86,16 +87,15 @@ my_linear_ND = LinearSurrogate(x, y, lb, ub)
 surrogate_optimize(objective_function_ND, SRBF(), lb, ub, my_linear_ND, SobolSample(),
     maxiters = 15)
 
-#=
 #SVM
-lb = [1.0,1.0]
-ub = [6.0,6.0]
-x = sample(5,lb,ub,SobolSample())
-objective_function_ND = z -> 3*norm(z)+1
+lb = [1.0, 1.0]
+ub = [6.0, 6.0]
+x = sample(5, lb, ub, SobolSample())
+objective_function_ND = z -> 3 * norm(z) + 1
 y = objective_function_ND.(x)
-my_SVM_ND = SVMSurrogate(x,y,lb,ub)
-surrogate_optimize(objective_function_ND,SRBF(),lb,ub,my_SVM_ND,SobolSample(),maxiters=15)
-=#
+my_SVM_ND = SVMSurrogate(x, y, lb, ub)
+surrogate_optimize(
+    objective_function_ND, SRBF(), lb, ub, my_SVM_ND, SobolSample(), maxiters = 15)
 
 #Inverse distance surrogate
 lb = [1.0, 1.0]
@@ -276,26 +276,24 @@ num_centers = 2
 surrogate_optimize(objective_function_ND, SOP(num_centers), lb, ub, my_k_SOPND,
     SobolSample(), maxiters = 20)
 
-#multi optimization
-#=
-f  = x -> [x^2, x]
+f = x -> [x^2, x]
 lb = 1.0
 ub = 10.0
-x  = sample(100, lb, ub, SobolSample())
-y  = f.(x)
+x = sample(100, lb, ub, SobolSample())
+y = f.(x)
 my_radial_basis_smb = RadialBasis(x, y, lb, ub, rad = linearRadial())
-surrogate_optimize(f,SMB(),lb,ub,my_radial_basis_ego,SobolSample())
+surrogate_optimize(f, SMB(), lb, ub, my_radial_basis_smb, SobolSample())
 
-f  = x -> [x^2, x]
+f = x -> [x, sin(x)]
 lb = 1.0
 ub = 10.0
-x  = sample(100, lb, ub, SobolSample())
-y  = f.(x)
+x = sample(500, lb, ub, RandomSample())
+y = f.(x)
 my_radial_basis_rtea = RadialBasis(x, y, lb, ub, rad = linearRadial())
 Z = 0.8 #percentage
 K = 2 #number of revaluations
 p_cross = 0.5 #crossing vs copy
 n_c = 1.0 # hyperparameter for children creation
 sigma = 1.5 # mutation
-surrogate_optimize(f,RTEA(Z,K,p_cross,n_c,sigma),lb,ub,my_radial_basis_rtea,SobolSample())
-=#
+surrogate_optimize(f, RTEA(K, Z, p_cross, n_c, sigma), lb, ub,
+    my_radial_basis_rtea, SobolSample(); maxiters = 10)