General updates

.github/workflows/CI.yml

@@ -14,6 +14,7 @@ jobs:
   test:
     runs-on: ubuntu-latest
     strategy:
+      fail-fast: false
       matrix:
         group:
           - All

Project.toml

@@ -16,16 +16,16 @@ SciMLBase = "0bca4576-84f4-4d90-8ffe-ffa030f20462"
 
 [compat]
 Calculus = "0.5"
-Dierckx = "0.4, 0.5"
+Dierckx = "0.5"
 DiffEqBase = "6"
 Distributions = "0.25"
 ForwardDiff = "0.10"
-PenaltyFunctions = "0.1, 0.2, 0.3"
-PreallocationTools = "0.2, 0.3, 0.4"
-RecursiveArrayTools = "1.0, 2.0, 3"
+PenaltyFunctions = "0.3"
+PreallocationTools = "0.4"
+RecursiveArrayTools = "3"
 SciMLBase = "1.69, 2"
 SciMLSensitivity = "7"
-julia = "1.6"
+julia = "1.10"
 
 [extras]
 BlackBoxOptim = "a134a8b2-14d6-55f6-9291-3336d3ab0209"
@@ -38,7 +38,6 @@ OptimizationBBO = "3e6eede4-6085-4f62-9a71-46d9bc1eb92b"
 OptimizationNLopt = "4e6fcdb7-1186-4e1f-a706-475e75c168bb"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
-ParameterizedFunctions = "65888b18-ceab-5e60-b2b9-181511a3b968"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 SteadyStateDiffEq = "9672c7b4-1e72-59bd-8a11-6ac3964bc41f"
@@ -48,4 +47,4 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [targets]
-test = ["Test", "BlackBoxOptim", "DelayDiffEq", "ForwardDiff", "NLopt", "Optim", "Optimization", "OptimizationBBO", "OptimizationNLopt", "OptimizationOptimJL", "OrdinaryDiffEq", "ParameterizedFunctions", "Random", "SciMLSensitivity", "StochasticDiffEq", "SteadyStateDiffEq", "Sundials", "Zygote"]
+test = ["Test", "BlackBoxOptim", "DelayDiffEq", "ForwardDiff", "NLopt", "Optim", "Optimization", "OptimizationBBO", "OptimizationNLopt", "OptimizationOptimJL", "OrdinaryDiffEq", "Random", "SciMLSensitivity", "StochasticDiffEq", "SteadyStateDiffEq", "Sundials", "Zygote"]

src/DiffEqParamEstim.jl

@@ -4,9 +4,13 @@ using DiffEqBase, PenaltyFunctions,
       LinearAlgebra, Dierckx, SciMLBase
 
 import PreallocationTools
-STANDARD_PROB_GENERATOR(prob, p) = remake(prob; u0 = eltype(p).(prob.u0), p = p)
+function STANDARD_PROB_GENERATOR(prob, p)
+    f = ODEFunction{isinplace(prob), SciMLBase.FullSpecialize}(SciMLBase.unwrapped_f(prob.f))
+    remake(prob; u0 = eltype(p).(prob.u0), p = p, f = f)
+end
 function STANDARD_PROB_GENERATOR(prob::EnsembleProblem, p)
-    EnsembleProblem(remake(prob.prob; u0 = eltype(p).(prob.prob.u0), p = p),
+    f = ODEFunction{isinplace(prob.prob), SciMLBase.FullSpecialize}(SciMLBase.unwrapped_f(prob.prob.f))
+    EnsembleProblem(remake(prob.prob; u0 = eltype(p).(prob.prob.u0), p = p, f = f),
         output_func = prob.output_func,
         prob_func = prob.prob_func,
         reduction = prob.reduction,
@@ -17,8 +21,9 @@ STANDARD_MS_PROB_GENERATOR = function (prob, p, k)
     P, N = length(prob.p), length(prob.u0)
     K = Int((length(p) - P) / N)
     τ = range(t0, tf, length = K + 1)
+    f = ODEFunction{isinplace(prob), SciMLBase.FullSpecialize}(SciMLBase.unwrapped_f(prob.f))
     remake(prob; u0 = p[(1 + (k - 1) * N):(k * N)], p = p[(end - P + 1):end],
-        tspan = (τ[k], τ[k + 1]))
+        tspan = (τ[k], τ[k + 1]), f = f)
 end
 
 include("cost_functions.jl")

test/multiple_shooting_objective_test.jl

@@ -28,5 +28,5 @@ ms_obj1 = multiple_shooting_objective(ms_prob, Tsit5(), L2Loss(t, data),
     reltol = 1e-6)
 optprob = Optimization.OptimizationProblem(ms_obj1, zeros(18), lb = first.(bound),
     ub = last.(bound))
-result = solve(optprob, BFGS(), maxiters = 500)
+result = solve(optprob, BFGS(), maxiters = 10000)
 @test result.u[(end - 1):end]≈[1.5, 1.0] atol=2e-1

test/tests_on_odes/blackboxoptim_test.jl

@@ -4,17 +4,17 @@ println("Use BlackBoxOptim to fit the parameter")
 cost_function = build_loss_objective(prob1, Tsit5(), L2Loss(t, data),
     maxiters = 10000, verbose = false)
 bound1 = Tuple{Float64, Float64}[(1, 2)]
-result = bboptimize(cost_function; SearchRange = bound1, MaxSteps = 11e3)
+result = bboptimize(cost_function; SearchRange = bound1, MaxSteps = 20e3)
 @test result.archive_output.best_candidate[1]≈1.5 atol=3e-1
 
-cost_function = build_loss_objective(prob2, Tsit5(), L2Loss(t, data),
+cost_function = build_loss_objective(prob2, Tsit5(), L2Loss(t, data2),
     maxiters = 10000, verbose = false)
 bound2 = Tuple{Float64, Float64}[(1, 2), (2, 4)]
-result = bboptimize(cost_function; SearchRange = bound2, MaxSteps = 11e3)
+result = bboptimize(cost_function; SearchRange = bound2, MaxSteps = 20e3)
 @test result.archive_output.best_candidate≈[1.5; 3.0] atol=3e-1
 
-cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data),
+cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data3),
     maxiters = 10000, verbose = false)
 bound3 = Tuple{Float64, Float64}[(1, 2), (0, 2), (2, 4), (0, 2)]
-result = bboptimize(cost_function; SearchRange = bound3, MaxSteps = 11e3)
+result = bboptimize(cost_function; SearchRange = bound3, MaxSteps = 100e3)
 @test result.archive_output.best_candidate≈[1.5; 1.0; 3.0; 1.0] atol=5e-1

test/tests_on_odes/genetic_algorithm_test.jl

@@ -25,7 +25,7 @@ result, fitness, cnt = ga(cost_function, N;
     mutation = domainrange(fill(0.5, N)))
 @test result[1]≈1.5 atol=3e-1
 
-cost_function = build_loss_objective(prob2, Tsit5(), L2Loss(t, data),
+cost_function = build_loss_objective(prob2, Tsit5(), L2Loss(t, data2),
     maxiters = 10000)
 N = 2
 result, fitness, cnt = ga(cost_function, N;
@@ -37,7 +37,7 @@ result, fitness, cnt = ga(cost_function, N;
     mutation = domainrange(fill(0.5, N)))
 @test result≈[1.5; 3.0] atol=3e-1
 
-cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data),
+cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data3),
     maxiters = 10000)
 N = 4
 result, fitness, cnt = ga(cost_function, N;

test/tests_on_odes/l2loss_test.jl

@@ -7,14 +7,14 @@ result = bboptimize(cost_function; SearchRange = bound1, MaxSteps = 11e3)
 @test result.archive_output.best_candidate[1]≈1.5 atol=3e-1
 
 cost_function = build_loss_objective(prob2, Tsit5(),
-    L2Loss(t, data, differ_weight = nothing,
+    L2Loss(t, data2, differ_weight = nothing,
         data_weight = 1.0),
     maxiters = 10000, verbose = false)
 bound2 = Tuple{Float64, Float64}[(1, 2), (1, 4)]
 result = bboptimize(cost_function; SearchRange = bound2, MaxSteps = 11e3)
 @test result.archive_output.best_candidate≈[1.5; 3.0] atol=3e-1
 
-cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data, differ_weight = 10),
+cost_function = build_loss_objective(prob3, Tsit5(), L2Loss(t, data3, differ_weight = 10),
     maxiters = 10000, verbose = false)
 bound3 = Tuple{Float64, Float64}[(1, 2), (0, 2), (2, 4), (0, 2)]
 result = bboptimize(cost_function; SearchRange = bound3, MaxSteps = 11e3)

test/tests_on_odes/nlopt_test.jl

@@ -22,7 +22,7 @@ opt = Opt(:GN_ISRES, 1)
 lower_bounds!(opt, [1.0])
 upper_bounds!(opt, [3.0])
 xtol_rel!(opt, 1e-4)
-maxeval!(opt, 100 - 000)
+maxeval!(opt, 10000)
 res = solve(optprob, opt)
 @test res.u[1]≈1.5 atol=1e-1
 

test/tests_on_odes/optim_test.jl

@@ -14,13 +14,13 @@ result = Optim.optimize(obj, [1.0], Optim.BFGS())
 #sol_optimized2 = solve(prob)
 #plot!(sol_optimized2,leg=false)
 
-cost_function2 = build_loss_objective(prob2, Tsit5(), L2Loss(t, data),
+cost_function2 = build_loss_objective(prob2, Tsit5(), L2Loss(t, data2),
     maxiters = 10000, verbose = false)
 result_bfgs = Optim.optimize(cost_function2, [1.0, 2.5], Optim.BFGS())
 @test_broken result_bfgs.minimizer≈[1.5; 3.0] atol=3e-1
 
 Random.seed!(200)
-cost_function3 = build_loss_objective(prob3, Tsit5(), L2Loss(t, data),
+cost_function3 = build_loss_objective(prob3, Tsit5(), L2Loss(t, data3),
     maxiters = 10000, verbose = false)
 result_bfgs = Optim.optimize(cost_function3, [1.3, 0.8, 2.8, 1.2], Optim.BFGS())
 @test result_bfgs.minimizer≈[1.5; 1.0; 3.0; 1.0] atol=5e-1

test/tests_on_odes/regularization_test.jl

@@ -4,14 +4,14 @@ cost_function_1 = build_loss_objective(prob1, Tsit5(), L2Loss(t, data),
     Optimization.AutoZygote(),
     Regularization(0.6, L2Penalty()), maxiters = 10000,
     verbose = false, abstol = 1e-8, reltol = 1e-8)
-cost_function_2 = build_loss_objective(prob2, Tsit5(), L2Loss(t, data),
+cost_function_2 = build_loss_objective(prob2, Tsit5(), L2Loss(t, data2),
     Optimization.AutoZygote(),
     Regularization(0.1,
         MahalanobisPenalty(Matrix(1.0I, 2, 2))),
     verbose = false,
     abstol = 1e-8, reltol = 1e-8,
     maxiters = 10000)
-cost_function_3 = build_loss_objective(prob3, Tsit5(), L2Loss(t, data),
+cost_function_3 = build_loss_objective(prob3, Tsit5(), L2Loss(t, data3),
     Optimization.AutoZygote(),
     Regularization(0.1,
         MahalanobisPenalty(Matrix(1.0I, 4, 4))),
@@ -20,14 +20,14 @@ cost_function_3 = build_loss_objective(prob3, Tsit5(), L2Loss(t, data),
     maxiters = 10000)
 
 println("Use Optim BFGS to fit the parameter")
-optprob = Optimization.OptimizationProblem(cost_function_1, [1.0])
+optprob = Optimization.OptimizationProblem(cost_function_1, [1.0], lb=[1.0], ub=[2.0])
 result = solve(optprob, Optim.BFGS())
 @test result.u[1]≈1.5 atol=3e-1
 
-optprob = Optimization.OptimizationProblem(cost_function_2, [1.2, 2.7])
+optprob = Optimization.OptimizationProblem(cost_function_2, [1.2, 2.7], lb=[1.0, 2.0], ub=[2.0, 4.0])
 result = solve(optprob, Optim.BFGS())
 @test result.minimizer≈[1.5; 3.0] atol=3e-1
 
-optprob = Optimization.OptimizationProblem(cost_function_3, [1.3, 0.8, 2.8, 1.2])
+optprob = Optimization.OptimizationProblem(cost_function_3, [1.3, 0.8, 2.8, 1.2], lb=[1.0, 0.5, 2.0, 0.5], ub=[2.0, 1.5, 4.0, 1.5])
 result = solve(optprob, Optim.BFGS())
 @test result.minimizer≈[1.5; 1.0; 3.0; 1.0] atol=5e-1

test/tests_on_odes/test_problems.jl

@@ -1,32 +1,54 @@
-using OrdinaryDiffEq, ParameterizedFunctions, RecursiveArrayTools
+using OrdinaryDiffEq, RecursiveArrayTools, Random
+
+# Set seed for reproducible test data
+Random.seed!(100)
 
 # Here are the problems to solve
 
-f1 = @ode_def begin
-    dx = a * x - x * y
-    dy = -3y + x * y
-end a
+function f1!(du, u, p, t)
+    a = p[1]
+    x, y = u
+    du[1] = a * x - x * y
+    du[2] = -3y + x * y
+end
+
 u0 = [1.0; 1.0]
 tspan = (0.0, 10.0)
 p = [1.5]
-prob1 = ODEProblem(f1, u0, tspan, [1.5])
+prob1 = ODEProblem(f1!, u0, tspan, [1.5])
+
+function f2!(du, u, p, t)
+    a, c = p
+    x, y = u
+    du[1] = a * x - x * y
+    du[2] = -c * y + x * y
+end
 
-f2 = @ode_def begin
-    dx = a * x - x * y
-    dy = -c * y + x * y
-end a c
 p = [1.5, 3.0]
-prob2 = ODEProblem(f2, u0, tspan, p)
+prob2 = ODEProblem(f2!, u0, tspan, p)
+
+function f3!(du, u, p, t)
+    a, b, c, d = p
+    x, y = u
+    du[1] = a * x - b * x * y
+    du[2] = -c * y + d * x * y
+end
 
-f3 = @ode_def begin
-    dx = a * x - b * x * y
-    dy = -c * y + d * x * y
-end a b c d
 p = [1.5, 1.0, 3.0, 1.0]
-prob3 = ODEProblem(f3, u0, tspan, p)
+prob3 = ODEProblem(f3!, u0, tspan, p)
 
 # Generate random data based off of the known solution
-sol = solve(prob1, Tsit5())
+sol1 = solve(prob1, Tsit5())
 t = collect(range(0, stop = 10, length = 200))
-randomized = VectorOfArray([(sol(t[i]) + 0.01randn(2)) for i in 1:length(t)])
-data = convert(Array, randomized)
+randomized1 = VectorOfArray([(sol1(t[i]) + 0.01randn(2)) for i in 1:length(t)])
+data = convert(Array, randomized1)
+
+# Generate data for prob2
+sol2 = solve(prob2, Tsit5())
+randomized2 = VectorOfArray([(sol2(t[i]) + 0.01randn(2)) for i in 1:length(t)])
+data2 = convert(Array, randomized2)
+
+# Generate data for prob3
+sol3 = solve(prob3, Tsit5())
+randomized3 = VectorOfArray([(sol3(t[i]) + 0.01randn(2)) for i in 1:length(t)])
+data3 = convert(Array, randomized3)

test/tests_on_odes/weighted_loss_test.jl

@@ -24,6 +24,9 @@ weighted_cost_function = build_loss_objective(prob1, Tsit5(),
         data_weight = weight),
     maxiters = 10000, verbose = false)
 opt = Opt(:LN_COBYLA, 1)
+lower_bounds!(opt, [1.0])
+upper_bounds!(opt, [2.0])
+maxeval!(opt, 10000)
 min_objective!(opt, weighted_cost_function)
 (minf, minx, ret) = NLopt.optimize(opt, [1.3])
 @test minx[1]≈1.5 atol=1e-2