Add bound constraints to the Formulation object

src/predictors/Affine.jl

@@ -22,7 +22,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2]);
+julia> @variable(model, 0 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.Affine([2.0 3.0], [4.0])
 Affine(A, b) [input: 2, output: 1]
@@ -37,7 +37,9 @@ julia> formulation
 Affine(A, b) [input: 2, output: 1]
 ├ variables [1]
 │ └ moai_Affine[1]
-└ constraints [1]
+└ constraints [3]
+  ├ moai_Affine[1] ≥ 4
+  ├ moai_Affine[1] ≤ 12
   └ 2 x[1] + 3 x[2] - moai_Affine[1] = -4
 
 julia> y, formulation =
@@ -75,6 +77,7 @@ function add_predictor(model::JuMP.AbstractModel, predictor::Affine, x::Vector)
     m = size(predictor.A, 1)
     y = JuMP.@variable(model, [1:m], base_name = "moai_Affine")
     bounds = _get_variable_bounds.(x)
+    cons = Any[]
     for i in 1:size(predictor.A, 1)
         y_lb, y_ub = predictor.b[i], predictor.b[i]
         for j in 1:size(predictor.A, 2)
@@ -83,9 +86,9 @@ function add_predictor(model::JuMP.AbstractModel, predictor::Affine, x::Vector)
             y_ub += a_ij * ifelse(a_ij >= 0, ub, lb)
             y_lb += a_ij * ifelse(a_ij >= 0, lb, ub)
         end
-        _set_bounds_if_finite(y[i], y_lb, y_ub)
+        _set_bounds_if_finite(cons, y[i], y_lb, y_ub)
     end
-    cons = JuMP.@constraint(model, predictor.A * x .+ predictor.b .== y)
+    append!(cons, JuMP.@constraint(model, predictor.A * x .+ predictor.b .== y))
     return y, Formulation(predictor, y, cons)
 end
 

src/predictors/Pipeline.jl

@@ -46,7 +46,9 @@ ReLUQuadratic()
 ├ variables [2]
 │ ├ moai_ReLU[1]
 │ └ moai_z[1]
-└ constraints [2]
+└ constraints [4]
+  ├ moai_ReLU[1] ≥ 0
+  ├ moai_z[1] ≥ 0
   ├ moai_Affine[1] - moai_ReLU[1] + moai_z[1] = 0
   └ moai_ReLU[1]*moai_z[1] = 0
 ```

src/predictors/ReLU.jl

@@ -17,7 +17,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2]);
+julia> @variable(model, -1 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.ReLU()
 ReLU()
@@ -34,9 +34,11 @@ ReLU()
 ├ variables [2]
 │ ├ moai_ReLU[1]
 │ └ moai_ReLU[2]
-└ constraints [4]
+└ constraints [6]
   ├ moai_ReLU[1] ≥ 0
+  ├ moai_ReLU[1] ≤ 1
   ├ moai_ReLU[2] ≥ 0
+  ├ moai_ReLU[2] ≤ 2
   ├ moai_ReLU[1] - max(0.0, x[1]) = 0
   └ moai_ReLU[2] - max(0.0, x[2]) = 0
 
@@ -57,12 +59,10 @@ ReducedSpace(ReLU())
 struct ReLU <: AbstractPredictor end
 
 function add_predictor(model::JuMP.AbstractModel, predictor::ReLU, x::Vector)
-    ub = last.(_get_variable_bounds.(x))
     y = JuMP.@variable(model, [1:length(x)], base_name = "moai_ReLU")
-    _set_bounds_if_finite.(y, 0, max.(0, ub))
-    cons = JuMP.@constraint(model, y .== max.(0, x))
-    constraints = Any[JuMP.LowerBoundRef.(y); cons]
-    return y, Formulation(predictor, y, constraints)
+    cons = _set_direct_bounds(x -> max(0, x), 0, nothing, x, y)
+    append!(cons, JuMP.@constraint(model, y .== max.(0, x)))
+    return y, Formulation(predictor, y, cons)
 end
 
 function add_predictor(
@@ -103,17 +103,21 @@ ReLUBigM(100.0)
 ├ variables [4]
 │ ├ moai_ReLU[1]
 │ ├ moai_ReLU[2]
-│ ├ _[5]
-│ └ _[6]
-└ constraints [8]
-  ├ _[5] binary
+│ ├ moai_z[1]
+│ └ moai_z[2]
+└ constraints [12]
+  ├ moai_ReLU[1] ≥ 0
+  ├ moai_ReLU[1] ≤ 1
+  ├ moai_ReLU[2] ≥ 0
+  ├ moai_ReLU[2] ≤ 2
+  ├ moai_z[1] binary
   ├ -x[1] + moai_ReLU[1] ≥ 0
-  ├ moai_ReLU[1] - _[5] ≤ 0
-  ├ -x[1] + moai_ReLU[1] + 3 _[5] ≤ 3
-  ├ _[6] binary
+  ├ moai_ReLU[1] - moai_z[1] ≤ 0
+  ├ -x[1] + moai_ReLU[1] + 3 moai_z[1] ≤ 3
+  ├ moai_z[2] binary
   ├ -x[2] + moai_ReLU[2] ≥ 0
-  ├ moai_ReLU[2] - 2 _[6] ≤ 0
-  └ -x[2] + moai_ReLU[2] + 3 _[6] ≤ 3
+  ├ moai_ReLU[2] - 2 moai_z[2] ≤ 0
+  └ -x[2] + moai_ReLU[2] + 3 moai_z[2] ≤ 3
 ```
 """
 struct ReLUBigM <: AbstractPredictor
@@ -126,14 +130,14 @@ function add_predictor(
     x::Vector,
 )
     m = length(x)
-    bounds = _get_variable_bounds.(x)
     y = JuMP.@variable(model, [1:m], base_name = "moai_ReLU")
-    _set_bounds_if_finite.(y, 0, max.(0, last.(bounds)))
-    formulation = Formulation(predictor)
+    cons = _set_direct_bounds(x -> max(0, x), 0, nothing, x, y)
+    formulation = Formulation(predictor, Any[], cons)
     append!(formulation.variables, y)
     for i in 1:m
-        lb, ub = bounds[i]
+        lb, ub = _get_variable_bounds(x[i])
         z = JuMP.@variable(model, binary = true)
+        JuMP.set_name(z, "moai_z[$i]")
         push!(formulation.variables, z)
         push!(formulation.constraints, JuMP.BinaryRef(z))
         c = JuMP.@constraint(model, y[i] >= x[i])
@@ -167,7 +171,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2] >= -1);
+julia> @variable(model, -1 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.ReLUSOS1()
 ReLUSOS1()
@@ -186,7 +190,13 @@ ReLUSOS1()
 │ ├ moai_ReLU[2]
 │ ├ moai_z[1]
 │ └ moai_z[2]
-└ constraints [4]
+└ constraints [10]
+  ├ moai_ReLU[1] ≥ 0
+  ├ moai_ReLU[1] ≤ 1
+  ├ moai_ReLU[2] ≥ 0
+  ├ moai_ReLU[2] ≤ 2
+  ├ moai_z[1] ≤ 1
+  ├ moai_z[2] ≤ 1
   ├ x[1] - moai_ReLU[1] + moai_z[1] = 0
   ├ x[2] - moai_ReLU[2] + moai_z[2] = 0
   ├ [moai_ReLU[1], moai_z[1]] ∈ MathOptInterface.SOS1{Float64}([1.0, 2.0])
@@ -203,11 +213,11 @@ function add_predictor(
     m = length(x)
     bounds = _get_variable_bounds.(x)
     y = JuMP.@variable(model, [i in 1:m], base_name = "moai_ReLU")
-    _set_bounds_if_finite.(y, 0, max.(0, last.(bounds)))
+    cons = _set_direct_bounds(x -> max(0, x), 0, nothing, x, y)
     z = JuMP.@variable(model, [1:m], lower_bound = 0, base_name = "moai_z")
-    _set_bounds_if_finite.(z, nothing, -first.(bounds))
-    cons = JuMP.@constraint(model, x .== y - z)
-    formulation = Formulation(predictor, Any[y; z], Any[cons;])
+    _set_bounds_if_finite.(Ref(cons), z, nothing, -first.(bounds))
+    append!(cons, JuMP.@constraint(model, x .== y - z))
+    formulation = Formulation(predictor, Any[y; z], cons)
     for i in 1:m
         c = JuMP.@constraint(model, [y[i], z[i]] in MOI.SOS1([1.0, 2.0]))
         push!(formulation.constraints, c)
@@ -235,7 +245,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2] >= -1);
+julia> @variable(model, -1 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.ReLUQuadratic()
 ReLUQuadratic()
@@ -254,7 +264,15 @@ ReLUQuadratic()
 │ ├ moai_ReLU[2]
 │ ├ moai_z[1]
 │ └ moai_z[2]
-└ constraints [4]
+└ constraints [12]
+  ├ moai_ReLU[1] ≥ 0
+  ├ moai_ReLU[1] ≤ 1
+  ├ moai_ReLU[2] ≥ 0
+  ├ moai_ReLU[2] ≤ 2
+  ├ moai_z[1] ≥ 0
+  ├ moai_z[1] ≤ 1
+  ├ moai_z[2] ≥ 0
+  ├ moai_z[2] ≤ 1
   ├ x[1] - moai_ReLU[1] + moai_z[1] = 0
   ├ x[2] - moai_ReLU[2] + moai_z[2] = 0
   ├ moai_ReLU[1]*moai_z[1] = 0
@@ -271,10 +289,10 @@ function add_predictor(
     m = length(x)
     bounds = _get_variable_bounds.(x)
     y = JuMP.@variable(model, [1:m], base_name = "moai_ReLU")
-    _set_bounds_if_finite.(y, 0, max.(0, last.(bounds)))
+    cons = _set_direct_bounds(x -> max(0, x), 0, nothing, x, y)
     z = JuMP.@variable(model, [1:m], base_name = "moai_z")
-    _set_bounds_if_finite.(z, 0, max.(0, -first.(bounds)))
-    c1 = JuMP.@constraint(model, x .== y - z)
-    c2 = JuMP.@constraint(model, y .* z .== 0)
-    return y, Formulation(predictor, Any[y; z], Any[c1; c2])
+    _set_bounds_if_finite.(Ref(cons), z, 0, max.(0, -first.(bounds)))
+    append!(cons, JuMP.@constraint(model, x .== y - z))
+    append!(cons, JuMP.@constraint(model, y .* z .== 0))
+    return y, Formulation(predictor, Any[y; z], cons)
 end

src/predictors/Scale.jl

@@ -22,7 +22,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2]);
+julia> @variable(model, 0 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.Scale([2.0, 3.0], [4.0, 5.0])
 Scale(scale, bias)
@@ -39,7 +39,11 @@ Scale(scale, bias)
 ├ variables [2]
 │ ├ moai_Scale[1]
 │ └ moai_Scale[2]
-└ constraints [2]
+└ constraints [6]
+  ├ moai_Scale[1] ≥ 4
+  ├ moai_Scale[1] ≤ 6
+  ├ moai_Scale[2] ≥ 5
+  ├ moai_Scale[2] ≤ 11
   ├ 2 x[1] - moai_Scale[1] = -4
   └ 3 x[2] - moai_Scale[2] = -5
 
@@ -70,14 +74,18 @@ function add_predictor(model::JuMP.AbstractModel, predictor::Scale, x::Vector)
     m = length(predictor.scale)
     y = JuMP.@variable(model, [1:m], base_name = "moai_Scale")
     bounds = _get_variable_bounds.(x)
+    cons = Any[]
     for (i, scale) in enumerate(predictor.scale)
         y_lb = y_ub = predictor.bias[i]
         lb, ub = bounds[i]
         y_ub += scale * ifelse(scale >= 0, ub, lb)
         y_lb += scale * ifelse(scale >= 0, lb, ub)
-        _set_bounds_if_finite(y[i], y_lb, y_ub)
+        _set_bounds_if_finite(cons, y[i], y_lb, y_ub)
     end
-    cons = JuMP.@constraint(model, predictor.scale .* x .+ predictor.bias .== y)
+    append!(
+        cons,
+        JuMP.@constraint(model, predictor.scale .* x .+ predictor.bias .== y),
+    )
     return y, Formulation(predictor, y, cons)
 end
 

src/predictors/Sigmoid.jl

@@ -17,7 +17,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2]);
+julia> @variable(model, -1 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.Sigmoid()
 Sigmoid()
@@ -35,10 +35,10 @@ Sigmoid()
 │ ├ moai_Sigmoid[1]
 │ └ moai_Sigmoid[2]
 └ constraints [6]
-  ├ moai_Sigmoid[1] ≥ 0
-  ├ moai_Sigmoid[2] ≥ 0
-  ├ moai_Sigmoid[1] ≤ 1
-  ├ moai_Sigmoid[2] ≤ 1
+  ├ moai_Sigmoid[1] ≥ 0.2689414213699951
+  ├ moai_Sigmoid[1] ≤ 0.7310585786300049
+  ├ moai_Sigmoid[2] ≥ 0.2689414213699951
+  ├ moai_Sigmoid[2] ≤ 0.8807970779778823
   ├ moai_Sigmoid[1] - (1.0 / (1.0 + exp(-x[1]))) = 0
   └ moai_Sigmoid[2] - (1.0 / (1.0 + exp(-x[2]))) = 0
 
@@ -60,10 +60,9 @@ struct Sigmoid <: AbstractPredictor end
 
 function add_predictor(model::JuMP.AbstractModel, predictor::Sigmoid, x::Vector)
     y = JuMP.@variable(model, [1:length(x)], base_name = "moai_Sigmoid")
-    _set_bounds_if_finite.(y, 0, 1)
-    cons = JuMP.@constraint(model, y .== 1 ./ (1 .+ exp.(-x)))
-    constraints = Any[JuMP.LowerBoundRef.(y); JuMP.UpperBoundRef.(y); cons]
-    return y, Formulation(predictor, y, constraints)
+    cons = _set_direct_bounds(x -> 1 / (1 + exp(-x)), 0, 1, x, y)
+    append!(cons, JuMP.@constraint(model, y .== 1 ./ (1 .+ exp.(-x))))
+    return y, Formulation(predictor, y, cons)
 end
 
 function add_predictor(

src/predictors/SoftMax.jl

@@ -37,8 +37,8 @@ SoftMax()
 │ └ moai_SoftMax[2]
 └ constraints [8]
   ├ moai_SoftMax[1] ≥ 0
-  ├ moai_SoftMax[2] ≥ 0
   ├ moai_SoftMax[1] ≤ 1
+  ├ moai_SoftMax[2] ≥ 0
   ├ moai_SoftMax[2] ≤ 1
   ├ moai_SoftMax_denom ≥ 0
   ├ moai_SoftMax_denom - (0.0 + exp(x[2]) + exp(x[1])) = 0
@@ -66,19 +66,14 @@ struct SoftMax <: AbstractPredictor end
 
 function add_predictor(model::JuMP.AbstractModel, predictor::SoftMax, x::Vector)
     y = JuMP.@variable(model, [1:length(x)], base_name = "moai_SoftMax")
-    _set_bounds_if_finite.(y, 0, 1)
+    cons = Any[]
+    _set_bounds_if_finite.(Ref(cons), y, 0, 1)
     denom = JuMP.@variable(model, base_name = "moai_SoftMax_denom")
     JuMP.set_lower_bound(denom, 0)
-    d_con = JuMP.@constraint(model, denom == sum(exp.(x)))
-    cons = JuMP.@constraint(model, y .== exp.(x) ./ denom)
-    constraints = [
-        JuMP.LowerBoundRef.(y)
-        JuMP.UpperBoundRef.(y)
-        JuMP.LowerBoundRef(denom)
-        d_con
-        cons
-    ]
-    return y, Formulation(predictor, [denom; y], constraints)
+    push!(cons, JuMP.LowerBoundRef(denom))
+    push!(cons, JuMP.@constraint(model, denom == sum(exp.(x))))
+    append!(cons, JuMP.@constraint(model, y .== exp.(x) ./ denom))
+    return y, Formulation(predictor, [denom; y], cons)
 end
 
 function add_predictor(

src/predictors/SoftPlus.jl

@@ -17,7 +17,7 @@ julia> using JuMP, MathOptAI
 
 julia> model = Model();
 
-julia> @variable(model, x[1:2]);
+julia> @variable(model, -1 <= x[i in 1:2] <= i);
 
 julia> f = MathOptAI.SoftPlus(; beta = 2.0)
 SoftPlus(2.0)
@@ -34,9 +34,11 @@ SoftPlus(2.0)
 ├ variables [2]
 │ ├ moai_SoftPlus[1]
 │ └ moai_SoftPlus[2]
-└ constraints [4]
-  ├ moai_SoftPlus[1] ≥ 0
-  ├ moai_SoftPlus[2] ≥ 0
+└ constraints [6]
+  ├ moai_SoftPlus[1] ≥ 0.0634640055214863
+  ├ moai_SoftPlus[1] ≤ 1.0634640055214863
+  ├ moai_SoftPlus[2] ≥ 0.0634640055214863
+  ├ moai_SoftPlus[2] ≤ 2.0090749639589047
   ├ moai_SoftPlus[1] - (log(1.0 + exp(2 x[1])) / 2.0) = 0
   └ moai_SoftPlus[2] - (log(1.0 + exp(2 x[2])) / 2.0) = 0
 
@@ -64,11 +66,11 @@ function add_predictor(
     predictor::SoftPlus,
     x::Vector,
 )
+    β = predictor.beta
     y = JuMP.@variable(model, [1:length(x)], base_name = "moai_SoftPlus")
-    _set_bounds_if_finite.(y, 0, nothing)
-    beta = predictor.beta
-    cons = JuMP.@constraint(model, y .== log.(1 .+ exp.(beta .* x)) ./ beta)
-    return y, Formulation(predictor, y, Any[JuMP.LowerBoundRef.(y); cons])
+    cons = _set_direct_bounds(xi -> log(1 + exp(β * xi)) / β, 0, nothing, x, y)
+    append!(cons, JuMP.@constraint(model, y .== log.(1 .+ exp.(β .* x)) ./ β))
+    return y, Formulation(predictor, y, cons)
 end
 
 function add_predictor(