Using SearchSpaces to solve #46

Project.toml

@@ -1,7 +1,7 @@
 name = "Metaheuristics"
 uuid = "bcdb8e00-2c21-11e9-3065-2b553b22f898"
 authors = ["Jesus Mejia <[email protected]>"]
-version = "3.2.16"
+version = "3.3.1"
 
 [deps]
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
@@ -11,13 +11,15 @@ Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Requires = "ae029012-a4dd-5104-9daa-d747884805df"
+SearchSpaces = "eb7571c6-2196-4f03-99b8-52a5a35b3163"
 SnoopPrecompile = "66db9d55-30c0-4569-8b51-7e840670fc0c"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
 [compat]
 Distances = "^0.10.7"
 JMcDM = "^0.7.1"
 Requires = "1"
+SearchSpaces = "^0.1.2"
 SnoopPrecompile = "1"
 julia = "^1.5"
 

src/algorithms/ABC/ABC.jl

@@ -82,7 +82,7 @@ function initialize!(
     options.parallel_evaluation &&
         error("ABC is not supporting parallel evaluation. Put `options.parallel_evaluation=false`")
 
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
     if options.f_calls_limit == 0
         options.f_calls_limit = 10000D
@@ -113,19 +113,19 @@ function update_state!(
         kargs...
     )
 
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
     fobj = problem.f
     bees = status.population
     Ne = parameters.Ne
     No = parameters.No
-    bounds = problem.bounds
-    a = view(bounds, 1,:)
-    b = view(bounds, 2,:)
+    bounds = problem.search_space
+    a = bounds.lb
+    b = bounds.ub
 
     employedPhase!(bees,problem,  Ne)
     outlookerPhase!(bees,problem, No)
 
-    @inline genPos(D=D, a=Array(a), b = Array(b)) = a + (b - a) .* rand(D)
+    @inline genPos(D=D, a=a, b=b) = a + (b - a) .* rand(D)
     best = chooseBest(bees, status.best_sol)
 
     status.f_calls += Ne + No + scoutPhase!(bees, problem, genPos, parameters.limit)

src/algorithms/ABC/bee_dynamics.jl

@@ -55,7 +55,7 @@ function updateBee!(bee, bee2, problem)
     v = ϕ*(bee.sol.x - bee2.sol.x)
 
     x_new = bee.sol.x + v
-    replace_with_random_in_bounds!(x_new, problem.bounds)
+    replace_with_random_in_bounds!(x_new, problem.search_space)
 
     new_sol = create_solution(x_new, problem)
 

src/algorithms/CCMO/CCMO.jl

@@ -150,10 +150,10 @@ function reproduction(status, parameters::CCMO{NSGA2}, problem)
     # selection
     _s = TournamentSelection(;N=base.N ÷ 2)
     # crossover
-    _c = SBX(η=base.η_cr, p = base.p_cr, bounds = problem.bounds)
+    _c = SBX(η=base.η_cr, p = base.p_cr, bounds = problem.search_space)
     # mutation
-    base.p_m < 0.0 && (base.p_m = 1/size(bounds,2)) 
-    _m = PolynomialMutation(η=base.η_m, p = base.p_m, bounds = problem.bounds)
+    base.p_m < 0.0 && (base.p_m = 1/getdim(problem)) 
+    _m = PolynomialMutation(η=base.η_m, p = base.p_m, bounds = problem.search_space)
 
     # population 1
     parent_mask = selection(population, _s, fitness=fitness1)

src/algorithms/CGSA/CGSA.jl

@@ -126,12 +126,13 @@ function initialize!(
 
     Rnorm = parameters.Rnorm
     N = parameters.N
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
     fobj = problem.f
 
 
     # bounds vectors
-    low, up = problem.bounds[1,:], problem.bounds[2,:]
+    low = problem.search_space.lb
+    up = problem.search_space.ub
 
     max_it = 500
     options.iterations = options.iterations == 0 ? max_it : options.iterations
@@ -209,7 +210,7 @@ function update_state!(
     # Checking allowable range. 
     # X = correctPop(X, low, up)
     for i = 1:N
-        x = reset_to_violated_bounds!(X[i,:], problem.bounds)
+        x = reset_to_violated_bounds!(X[i,:], problem.search_space)
         X[i,:] = x
     end
 

src/algorithms/DE/DE.jl

@@ -155,7 +155,7 @@ function initialize!(
         args...;
         kargs...
     )
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
 
 
@@ -213,7 +213,7 @@ function reproduction(status, parameters::AbstractDifferentialEvolution, problem
         x = get_position(population[i])
         u = DE_mutation(population, F, strategy, 1)
         v = DE_crossover(x, u, CR)
-        evo_boundary_repairer!(v, xBest, problem.bounds)
+        evo_boundary_repairer!(v, xBest, problem.search_space)
         X[i,:] = v
     end
 

src/algorithms/ECA/CECA.jl

@@ -22,7 +22,7 @@ function update_state!(
     feasible_solutions = findall( s->s.is_feasible, population )
     weights = compute_weights(population)
 
-    X_next = zeros(parameters.N, size(problem.bounds, 2))
+    X_next = zeros(parameters.N, getdim(problem))
 
     # For each elements in Population
     for i = 1:parameters.N
@@ -52,7 +52,7 @@ function update_state!(
         mask = rand(length(y)) .< 1.0 / length(y)
         y[mask] = v[mask]
 
-        evo_boundary_repairer!(y, c, problem.bounds)
+        evo_boundary_repairer!(y, c, problem.search_space)
         X_next[i,:] = y
     end
 

src/algorithms/ECA/ECA.jl

@@ -119,7 +119,7 @@ function eca_solution(status, parameters, options, problem, I, i)
     end
     # binary crossover
     y, M_current = crossover(U[u_best].x, y, parameters.p_cr)
-    evo_boundary_repairer!(y, c, problem.bounds)
+    evo_boundary_repairer!(y, c, problem.search_space)
     y, M_current
 end
 
@@ -135,7 +135,7 @@ function update_state!(
     )
 
     I = randperm(parameters.N)
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
     parameters.adaptive && (Mcr_fail = zeros(D))
 
@@ -237,7 +237,7 @@ function initialize!(
         args...;
         kargs...
     )
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
 
     if parameters.N <= parameters.K
@@ -299,7 +299,7 @@ function reproduction(status, parameters::ECA, problem)
                               parameters.K,
                               parameters.η_max;
                               i=i,
-                             bounds = problem.bounds)
+                             bounds = problem.search_space)
     end
 
     X 

src/algorithms/ECA/center_of_mass.jl

@@ -141,7 +141,7 @@ function ECA_operator(
         return y
     end
 
-    !isempty(bounds) && evo_boundary_repairer!(y, c, bounds)
+    evo_boundary_repairer!(y, c, bounds)
 
     return y
 end

src/algorithms/MCCGA/MCCGA.jl

@@ -120,8 +120,8 @@ function initialize!(
         options.f_calls_limit = options.iterations * parameters.N + 1
     end
 
-    lower = problem.bounds[1,:]
-    upper = problem.bounds[2,:]
+    lower = problem.search_space.lb
+    upper = problem.search_space.ub
 
     parameters.probvector = initialprobs(lower, upper, maxsamples = parameters.maxsamples)
 

src/algorithms/MOEAD_DE/MOEAD_DE.jl

@@ -166,7 +166,7 @@ function initialize!(
 
 
     status = gen_initial_state(problem,parameters,information,options,status)
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
     parameters.nobjectives = length(status.population[1].f)
     parameters.p_m = parameters.p_m < 0.0 ? 1.0 / D : parameters.p_m
 
@@ -193,7 +193,7 @@ function update_state!(
     F = parameters.F
     CR = parameters.CR
 
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
 
     N = parameters.N
@@ -211,7 +211,7 @@ function update_state!(
         # reproduction
         v = MOEAD_DE_reproduction(i, P_idx, population, parameters, problem)
         # repair
-        replace_with_random_in_bounds!(v, problem.bounds)
+        replace_with_random_in_bounds!(v, problem.search_space)
         h = create_solution(v, problem)
         # update z
         update_reference_point!(parameters.z, h)
@@ -252,13 +252,13 @@ Perform Differential Evolution operators and polynomial mutation using three vec
 `a, b, c` and parameters `F, CR, p_m, η`, i.e., stepsize, crossover and
 mutation probability.
 """
-function MOEAD_DE_reproduction(a, b, c, F, CR, p_m, η, bounds)
+function MOEAD_DE_reproduction(a, b, c, F, CR, p_m, η, bounds::Bounds)
     D = length(a)
     # binomial crossover
     v = zeros(length(a))
 
-    la = view(bounds, 1, :)
-    lb = view(bounds, 2, :)
+    la = bounds.lb
+    lb = bounds.ub
 
     # binomial crossover
     for j in 1:D
@@ -309,7 +309,7 @@ function MOEAD_DE_reproduction(i, P_idx, population, parameters::MOEAD_DE, probl
                           parameters.CR,
                           parameters.p_m,
                           parameters.η,
-                          problem.bounds)
+                          problem.search_space)
 end
 
 g_te_ap(gx, V, τ, s1, s2) = V < τ ? gx + s1*V^2 : gx + s1*τ^2 + s2*(V - τ)

src/algorithms/NSGA2/NSGA2.jl

@@ -120,7 +120,7 @@ offspring. Return two vectors.
 """
 function GA_reproduction(pa::AbstractVector{T},
         pb::AbstractVector{T},
-        bounds;
+        bounds::Bounds;
         η_cr = 20,
         η_m  = 15,
         p_cr = 0.9,
@@ -156,7 +156,7 @@ Same that `GA_reproduction` but only returns one offspring.
 """
 function GA_reproduction_half(pa::AbstractVector{T},
         pb::AbstractVector{T},
-        bounds;
+        bounds::Bounds;
         η_cr = 20,
         η_m  = 15,
         p_cr = 0.9,
@@ -180,7 +180,7 @@ function reproduction(pa, pb, parameters::AbstractNSGA, problem)
     # crossover and mutation
     c1, c2 = GA_reproduction(get_position(pa),
                              get_position(pb),
-                             problem.bounds;
+                             problem.search_space;
                              η_cr = parameters.η_cr,
                              p_cr = parameters.p_cr,
                              η_m = parameters.η_m,
@@ -203,7 +203,7 @@ function initialize!(
     args...;
     kargs...
 )
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
     if parameters.p_m < 0.0
         parameters.p_m = 1.0 / D
@@ -256,15 +256,15 @@ function reproduction(status, parameters::AbstractNSGA, problem)
     @assert !isempty(status.population)
 
     N_half = parameters.N
-    Q = zeros(2N_half, size(problem.bounds, 2))
+    Q = zeros(2N_half, getdim(problem))
 
     for i in 1:N_half
         pa = tournament_selection(status.population)
         pb = tournament_selection(status.population)
 
         c1, c2 = GA_reproduction(get_position(pa),
                                  get_position(pb),
-                                 problem.bounds;
+                                 problem.search_space;
                                  η_cr = parameters.η_cr,
                                  p_cr = parameters.p_cr,
                                  η_m = parameters.η_m,

src/algorithms/NSGA3/NSGA3.jl

@@ -272,8 +272,9 @@ function initialize!(
     options::Options,
     args...;
     kargs...
-)
-    D = size(problem.bounds, 2)
+    )
+
+    D = getdim(problem)
 
     if parameters.p_m < 0.0
         parameters.p_m = 1.0 / D
@@ -324,15 +325,15 @@ function reproduction(status, parameters::NSGA3, problem)
     @assert !isempty(status.population)
 
     I = randperm(parameters.N)
-    Q = zeros(parameters.N, size(problem.bounds, 2))
+    Q = zeros(parameters.N, getdim(problem))
     for i = 1:parameters.N ÷ 2
 
         pa = status.population[I[2i-1]]
         pb = status.population[I[2i]]
 
         c1, c2 = GA_reproduction(get_position(pa),
                                  get_position(pb),
-                                 problem.bounds;
+                                 problem.search_space;
                                  η_cr = parameters.η_cr,
                                  p_cr = parameters.p_cr,
                                  η_m = parameters.η_m,

src/algorithms/PSO/PSO.jl

@@ -80,15 +80,15 @@ function update_state!(
     )
     xGBest = get_position(status.best_sol)
 
-    X_new = zeros(parameters.N, size(problem.bounds, 2))
+    X_new = zeros(parameters.N, getdim(problem))
 
     # For each elements in population
     for i in 1:parameters.N
         x = get_position(parameters.flock[i])
         xPBest = get_position(status.population[i])
         parameters.v[i, :] = velocity(x, parameters.v[i, :], xPBest, xGBest, parameters)
         x += parameters.v[i, :]
-        reset_to_violated_bounds!(x, problem.bounds)
+        reset_to_violated_bounds!(x, problem.search_space)
         X_new[i,:] = x
     end
 
@@ -119,7 +119,7 @@ function initialize!(
     args...;
     kargs...
        )
-    D = size(problem.bounds, 2)
+    D = getdim(problem)
 
 
     if parameters.N == 0

src/algorithms/SA/SA.jl

@@ -91,8 +91,8 @@ function initialize!(
    )
 
 
-    l = view( problem.bounds, 1,:)
-    u = view( problem.bounds, 2,:)
+    l = problem.search_space.lb
+    u = problem.search_space.ub
 
     if isempty(parameters.x_initial)
         parameters.x_initial = l .+ (u .- l) .* rand(length(u))
@@ -139,8 +139,8 @@ function update_state!(
     T = nevals / max_evals 
     μ = 10.0 ^( 100T )    
 
-    l = view( problem.bounds, 1,:)
-    u = view( problem.bounds, 2,:)
+    l = problem.search_space.lb
+    u = problem.search_space.ub
 
     # For each temperature we take 500 test points to simulate reach termal
     # equilibrium.
@@ -153,7 +153,7 @@ function update_state!(
 
         # Next step is to keep solution within bounds
         #x1 = (x1 .< l).*l+(l .<= x1).*(x1 .<= u).*x1+(u .< x1).*u			
-        reset_to_violated_bounds!(x1, problem.bounds)
+        reset_to_violated_bounds!(x1, problem.search_space)
         fx1 = evaluate(x1, problem)
 
         status.f_calls += 1