Support MOI.TimeLimitSec (#54)

Author: odow

README.md

@@ -67,3 +67,4 @@ the solution process.
  * `MOA.ObjectiveRelativeTolerance(index::Int)`
  * `MOA.ObjectiveWeight(index::Int)`
  * `MOA.SolutionLimit()`
+ * `MOI.TimeLimitSec()`

src/MultiObjectiveAlgorithms.jl

@@ -6,9 +6,7 @@
 module MultiObjectiveAlgorithms
 
 import Combinatorics
-import MathOptInterface
-
-const MOI = MathOptInterface
+import MathOptInterface as MOI
 
 struct SolutionPoint
     x::Dict{MOI.VariableIndex,Float64}
@@ -110,6 +108,7 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
     f::Union{Nothing,MOI.AbstractVectorFunction}
     solutions::Vector{SolutionPoint}
     termination_status::MOI.TerminationStatusCode
+    time_limit_sec::Union{Nothing,Float64}
 
     function Optimizer(optimizer_factory)
         return new(
@@ -118,6 +117,7 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
             nothing,
             SolutionPoint[],
             MOI.OPTIMIZE_NOT_CALLED,
+            nothing,
         )
     end
 end
@@ -143,6 +143,34 @@ function MOI.copy_to(dest::Optimizer, src::MOI.ModelLike)
     return MOI.Utilities.default_copy_to(dest, src)
 end
 
+### TimeLimitSec
+
+function MOI.supports(model::Optimizer, attr::MOI.TimeLimitSec)
+    return MOI.supports(model.inner, attr)
+end
+
+MOI.get(model::Optimizer, ::MOI.TimeLimitSec) = model.time_limit_sec
+
+function MOI.set(model::Optimizer, ::MOI.TimeLimitSec, value::Real)
+    model.time_limit_sec = Float64(value)
+    return
+end
+
+function MOI.set(model::Optimizer, ::MOI.TimeLimitSec, ::Nothing)
+    model.time_limit_sec = nothing
+    return
+end
+
+function _time_limit_exceeded(model::Optimizer, start_time::Float64)
+    time_limit = MOI.get(model, MOI.TimeLimitSec())
+    if time_limit === nothing
+        return false
+    end
+    return time() - start_time >= time_limit
+end
+
+### ObjectiveFunction
+
 function MOI.supports(
     ::Optimizer,
     ::MOI.ObjectiveFunction{<:MOI.AbstractScalarFunction},

src/algorithms/Chalmet.jl

@@ -11,6 +11,11 @@
 Chalmet, L.G., and Lemonidis, L., and Elzinga, D.J. (1986). An algorithm for the
 bi-criterion integer programming problem. European Journal of Operational
 Research. 25(2), 292-300
+
+## Supported optimizer attributes
+
+ * `MOI.TimeLimitSec()`: terminate if the time limit is exceeded and return the
+   list of current solutions.
 """
 mutable struct Chalmet <: AbstractAlgorithm end
 
@@ -37,6 +42,7 @@ function _solve_constrained_model(
 end
 
 function optimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
+    start_time = time()
     if MOI.output_dimension(model.f) != 2
         error("Chalmet requires exactly two objectives")
     end
@@ -91,6 +97,9 @@ function optimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
     push!(Q, (1, 2))
     t = 3
     while !isempty(Q)
+        if _time_limit_exceeded(model, start_time)
+            return MOI.TIME_LIMIT, solutions
+        end
         r, s = pop!(Q)
         yr, ys = solutions[r].y, solutions[s].y
         rhs = [max(yr[1], ys[1]), max(yr[2], ys[2])]

src/algorithms/Dichotomy.jl

@@ -13,7 +13,10 @@ Science 25(1), 73-78.
 
 ## Supported optimizer attributes
 
- * `MOA.SolutionLimit()`
+ * `MOI.TimeLimitSec()`: terminate if the time limit is exceeded and return the
+   list of current solutions.
+
+ * `MOA.SolutionLimit()`: terminate once this many solutions have been found.
 """
 mutable struct Dichotomy <: AbstractAlgorithm
     solution_limit::Union{Nothing,Int}
@@ -73,6 +76,7 @@ function _solve_weighted_sum(
 end
 
 function optimize_multiobjective!(algorithm::Dichotomy, model::Optimizer)
+    start_time = time()
     if MOI.output_dimension(model.f) > 2
         error("Only scalar or bi-objective problems supported.")
     end
@@ -93,7 +97,12 @@ function optimize_multiobjective!(algorithm::Dichotomy, model::Optimizer)
         push!(queue, (0.0, 1.0))
     end
     limit = MOI.get(algorithm, SolutionLimit())
+    status = MOI.OPTIMAL
     while length(queue) > 0 && length(solutions) < limit
+        if _time_limit_exceeded(model, start_time)
+            status = MOI.TIME_LIMIT
+            break
+        end
         (a, b) = popfirst!(queue)
         y_d = solutions[a].y .- solutions[b].y
         w = y_d[2] / (y_d[2] - y_d[1])
@@ -114,5 +123,5 @@ function optimize_multiobjective!(algorithm::Dichotomy, model::Optimizer)
     end
     solution_list =
         [solutions[w] for w in sort(collect(keys(solutions)); rev = true)]
-    return MOI.OPTIMAL, solution_list
+    return status, solution_list
 end

src/algorithms/DominguezRios.jl

@@ -11,6 +11,11 @@
 Dominguez-Rios, M.A. & Chicano, F., & Alba, E. (2021). Effective anytime
 algorithm for multiobjective combinatorial optimization problems. Information
 Sciences, 565(7), 210-228.
+
+## Supported optimizer attributes
+
+ * `MOI.TimeLimitSec()`: terminate if the time limit is exceeded and return the
+   list of current solutions.
 """
 mutable struct DominguezRios <: AbstractAlgorithm end
 
@@ -138,6 +143,7 @@ function _update!(
 end
 
 function optimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
+    start_time = time()
     sense = MOI.get(model.inner, MOI.ObjectiveSense())
     if sense == MOI.MAX_SENSE
         old_obj, neg_obj = copy(model.f), -model.f
@@ -187,7 +193,12 @@ function optimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
     t_max = MOI.add_variable(model.inner)
     solutions = SolutionPoint[]
     k = 0
+    status = MOI.OPTIMAL
     while any(!isempty(l) for l in L)
+        if _time_limit_exceeded(model, start_time)
+            status = MOI.TIME_LIMIT
+            break
+        end
         i, k = _select_next_box(L, k)
         B = L[k][i]
         w = 1 ./ max.(1, B.u - yI)
@@ -214,5 +225,5 @@ function optimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
         MOI.delete.(model.inner, constraints)
     end
     MOI.delete(model.inner, t_max)
-    return MOI.OPTIMAL, solutions
+    return status, solutions
 end

src/algorithms/EpsilonConstraint.jl

@@ -69,6 +69,7 @@ function optimize_multiobjective!(
     algorithm::EpsilonConstraint,
     model::Optimizer,
 )
+    start_time = time()
     if MOI.output_dimension(model.f) != 2
         error("EpsilonConstraint requires exactly two objectives")
     end
@@ -104,7 +105,12 @@ function optimize_multiobjective!(
         MOI.GreaterThan{Float64}, left
     end
     ci = MOI.add_constraint(model, f1, SetType(bound))
+    status = MOI.OPTIMAL
     while true
+        if _time_limit_exceeded(model, start_time)
+            status = MOI.TIME_LIMIT
+            break
+        end
         MOI.set(model, MOI.ConstraintSet(), ci, SetType(bound))
         MOI.optimize!(model.inner)
         if !_is_scalar_status_optimal(model)
@@ -121,5 +127,5 @@ function optimize_multiobjective!(
         end
     end
     MOI.delete(model, ci)
-    return MOI.OPTIMAL, filter_nondominated(sense, solutions)
+    return status, filter_nondominated(sense, solutions)
 end

src/algorithms/KirlikSayin.jl

@@ -16,6 +16,11 @@ This is an algorithm to generate all nondominated solutions for multi-objective
 discrete optimization problems. The algorithm maintains `(p-1)`-dimensional
 rectangle regions in the solution space, and a two-stage optimization problem
 is solved for each rectangle.
+
+## Supported optimizer attributes
+
+ * `MOI.TimeLimitSec()`: terminate if the time limit is exceeded and return the
+    list of current solutions.
 """
 mutable struct KirlikSayin <: AbstractAlgorithm end
 
@@ -74,6 +79,7 @@ end
 _volume(r::_Rectangle, l::Vector{Float64}) = prod(r.u - l)
 
 function optimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
+    start_time = time()
     sense = MOI.get(model.inner, MOI.ObjectiveSense())
     if sense == MOI.MAX_SENSE
         old_obj, neg_obj = copy(model.f), -model.f
@@ -134,7 +140,12 @@ function optimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
         MOI.LessThan{Float64},
         MOI.GreaterThan{Float64},
     )
+    status = MOI.OPTIMAL
     while !isempty(L)
+        if _time_limit_exceeded(model, start_time)
+            status = MOI.TIME_LIMIT
+            break
+        end
         Rᵢ = L[argmax([_volume(Rᵢ, _project(yI, k)) for Rᵢ in L])]
         lᵢ, uᵢ = Rᵢ.l, Rᵢ.u
         # Solving the first stage model: P_k(ε)
@@ -182,5 +193,5 @@ function optimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
         end
         _remove_rectangle(L, _Rectangle(Y_proj, uᵢ))
     end
-    return MOI.OPTIMAL, solutions
+    return status, solutions
 end

test/algorithms/Chalmet.jl

@@ -121,6 +121,43 @@ function test_knapsack_max()
     return
 end
 
+function test_time_limit()
+    n = 10
+    W = 2137.0
+    C = Float64[
+        566 611 506 180 817 184 585 423 26 317
+        62 84 977 979 874 54 269 93 881 563
+    ]
+    w = Float64[557, 898, 148, 63, 78, 964, 246, 662, 386, 272]
+    model = MOA.Optimizer(HiGHS.Optimizer)
+    MOI.set(model, MOA.Algorithm(), MOA.Chalmet())
+    MOI.set(model, MOI.Silent(), true)
+    MOI.set(model, MOI.TimeLimitSec(), 0.0)
+    x = MOI.add_variables(model, n)
+    MOI.add_constraint.(model, x, MOI.ZeroOne())
+    MOI.add_constraint(
+        model,
+        MOI.ScalarAffineFunction(
+            [MOI.ScalarAffineTerm(w[j], x[j]) for j in 1:n],
+            0.0,
+        ),
+        MOI.LessThan(W),
+    )
+    f = MOI.VectorAffineFunction(
+        [
+            MOI.VectorAffineTerm(i, MOI.ScalarAffineTerm(C[i, j], x[j])) for
+            i in 1:2 for j in 1:n
+        ],
+        [0.0, 0.0],
+    )
+    MOI.set(model, MOI.ObjectiveSense(), MOI.MAX_SENSE)
+    MOI.set(model, MOI.ObjectiveFunction{typeof(f)}(), f)
+    MOI.optimize!(model)
+    @test MOI.get(model, MOI.TerminationStatus()) == MOI.TIME_LIMIT
+    @test MOI.get(model, MOI.ResultCount()) > 0
+    return
+end
+
 function test_unbounded()
     model = MOA.Optimizer(HiGHS.Optimizer)
     MOI.set(model, MOA.Algorithm(), MOA.Chalmet())

test/algorithms/Dichotomy.jl

@@ -235,6 +235,37 @@ function test_biobjective_knapsack()
     return
 end
 
+function test_time_limit()
+    p1 = [77, 94, 71, 63, 96, 82, 85, 75, 72, 91, 99, 63, 84, 87, 79, 94, 90]
+    p2 = [65, 90, 90, 77, 95, 84, 70, 94, 66, 92, 74, 97, 60, 60, 65, 97, 93]
+    w = [80, 87, 68, 72, 66, 77, 99, 85, 70, 93, 98, 72, 100, 89, 67, 86, 91]
+    f = MOI.OptimizerWithAttributes(
+        () -> MOA.Optimizer(HiGHS.Optimizer),
+        MOA.Algorithm() => MOA.Dichotomy(),
+    )
+    model = MOI.instantiate(f)
+    MOI.set(model, MOI.Silent(), true)
+    MOI.set(model, MOI.TimeLimitSec(), 0.0)
+    x = MOI.add_variables(model, length(w))
+    MOI.add_constraint.(model, x, MOI.ZeroOne())
+    MOI.set(model, MOI.ObjectiveSense(), MOI.MAX_SENSE)
+    f = MOI.Utilities.operate(
+        vcat,
+        Float64,
+        [sum(1.0 * p[i] * x[i] for i in 1:length(w)) for p in [p1, p2]]...,
+    )
+    MOI.set(model, MOI.ObjectiveFunction{typeof(f)}(), f)
+    MOI.add_constraint(
+        model,
+        sum(1.0 * w[i] * x[i] for i in 1:length(w)),
+        MOI.LessThan(900.0),
+    )
+    MOI.optimize!(model)
+    @test MOI.get(model, MOI.TerminationStatus()) == MOI.TIME_LIMIT
+    @test MOI.get(model, MOI.ResultCount()) > 0
+    return
+end
+
 function test_infeasible()
     model = MOA.Optimizer(HiGHS.Optimizer)
     MOI.set(model, MOA.Algorithm(), MOA.Dichotomy())

test/algorithms/DominguezRios.jl

@@ -544,6 +544,45 @@ function test_no_bounding_box()
     return
 end
 
+function test_time_limit()
+    p = 3
+    n = 10
+    W = 2137.0
+    C = Float64[
+        566 611 506 180 817 184 585 423 26 317
+        62 84 977 979 874 54 269 93 881 563
+        664 982 962 140 224 215 12 869 332 537
+    ]
+    w = Float64[557, 898, 148, 63, 78, 964, 246, 662, 386, 272]
+    model = MOA.Optimizer(HiGHS.Optimizer)
+    MOI.set(model, MOA.Algorithm(), MOA.DominguezRios())
+    MOI.set(model, MOI.TimeLimitSec(), 0.0)
+    MOI.set(model, MOI.Silent(), true)
+    x = MOI.add_variables(model, n)
+    MOI.add_constraint.(model, x, MOI.ZeroOne())
+    MOI.add_constraint(
+        model,
+        MOI.ScalarAffineFunction(
+            [MOI.ScalarAffineTerm(w[j], x[j]) for j in 1:n],
+            0.0,
+        ),
+        MOI.LessThan(W),
+    )
+    f = MOI.VectorAffineFunction(
+        [
+            MOI.VectorAffineTerm(i, MOI.ScalarAffineTerm(-C[i, j], x[j]))
+            for i in 1:p for j in 1:n
+        ],
+        fill(0.0, p),
+    )
+    MOI.set(model, MOI.ObjectiveSense(), MOI.MIN_SENSE)
+    MOI.set(model, MOI.ObjectiveFunction{typeof(f)}(), f)
+    MOI.optimize!(model)
+    @test MOI.get(model, MOI.TerminationStatus()) == MOI.TIME_LIMIT
+    @test MOI.get(model, MOI.ResultCount()) == 0
+    return
+end
+
 end
 
 TestDominguezRios.run_tests()