Sei sulla pagina 1di 39

Solving Chance-Constrained Optimization Problems with

Stochastic Quadratic Inequalities


Miguel A. Lejeune∗, F. Margot†

Abstract
We study a complex class of stochastic programming problems involving a joint chance constraint
with random technology matrix and stochastic quadratic inequalities. We present a basic mixed-
integer nonlinear reformulation based on Boolean modeling and derive several variants of it. We
present detailed empirical results comparing the various reformulations and several easy to imple-
ment algorithmic ideas that improve performances of the mixed-integer nonlinear solver Couenne
for solving these problems. Guidelines on how to tune the solver and selecting reformulations are
presented. The test instances are epidemiology and disaster management facility location models and
cover the three types of stochastic quadratic inequalities, namely product of two variables that are (i)
both binary, (ii) binary and continuous, or (iii) both continuous.
Keywords: Stochastic Programming, Mixed-Integer Nonlinear Programming, Boolean Program-
ming, Nonlinear Branch-and-Bound Algorithm, Quadratic Stochastic Inequality, Joint Probabilistic
Constraint, Random Technology Matrix, Epidemiology, Facility Location

1 Introduction
A chance-constrained problem is a stochastic programming optimization problem involving one or more
stochastic constraints that must hold simultaneously with a minimum given probability. Applications of
chance constraint problems are numerous, e.g., management of resources [62], financial risk management
[14], process engineering [24] (see [53] for additional examples and references).
Chance constraint problems are very challenging to solve in practice as they typically involve non-
linearities, nonconvexities, discrete variables, and random variables. Even checking the feasibility of a
given solution might already be hard in practice and NP-hard in theory. In this paper, we assume no par-
ticular form for the joint distribution of the random variables. We consider that the distribution is given as
a collection of joint realizations of the random variables, i.e., a problem with a discrete joint distribution.
Probabilistic programming problems with finitely distributed random variables (see, e.g., [21, 52, 58])
are challenging and pervasive. Discrete distributions can capture features such as skewness or kurtosis,
and are frequently used either directly or as empirical approximations of a general distribution.
The problems studied in this paper are significantly more complex than typical chance constraint
problems studied in the literature. They have three distinguishing features: (i) the presence of several
stochastic inequalities that must be satisfied jointly with a specified probability (as opposed to having a
single stochastic inequality); (ii) the presence of random variables in the left-hand side of the inequalities
(i.e., as opposed to only in the right-hand side); (iii) the stochastic inequalities are quadratic, i.e., sums
of terms that may involve the product of up to two decision variables and one random variable (as
opposed to linear, where only sums of products of a single decision variable with a random variable can
occur). Chance constraint problems with features (i) and (ii) are known as having a multi-row random

George Washington University, Washington, DC, USA; mlejeune@gwu.edu

Carnegie Mellon University, Pittsburgh, PA, USA; fmargot@andrew.cmu.edu Supported by ONR grant N00014-12-10032.

1
technology matrix. We call problems with all three features Quadratic and Multi-row chance-constrained
optimization problems and denote them by QM.
In this paper, we develop theory and algorithmic approaches to solve QM to optimality. Problem
QM has r decision variables that can be continuous or binary. We assume that x j ∈ R for j = 1, 2, . . . , r1
and x j ∈ {0, 1} for j = r1 + 1, r1 + 2, . . . , r1 + r2 = r. In addition, it also has a set of random variables
ξ following a joint discrete distribution with finite support and not assumed to be independent. The
objective function is linear in the decision variables and does not involve the random variables. The
problem has a set of stochastic inequalities (indexed by i ∈ I) that must hold jointly with a specified
probability level p. Each such inequality includes products of two decision variables (x j1 and x j2 ) and
one stochastic variable ξi j1 j2 , and its right hand side is deterministic. The multi-row technology matrix is
therefore random. Let J be the set of all pairs of indices of decision variables appearing in a monomial
of these stochastic constraints. The problem might have additional deterministic constraints gb (x) ≤ 0
for b = 1, 2, . . . , m that are convex. The notation P refers to a probability measure.
The base formulation of problem QM thus reads:

QM : max qT x (1)
subject to gb (x) ≤ 0 , b = 1, . . . , m (2)
!
si j x j1 x j2 ξi j ≤ di , i ∈ I ≥ p
P
P (3)
j=( j1 , j2 )∈J
x ∈ Rr+1 × {0, 1}r2 . (4)

The constraints (3) are known as multi-row probabilistic constraint with random technology matrix.
For discretely distributed random variables, the feasible set of QM is non-convex even when all the deci-
sion variables are continuous and the deterministic constraints are linear [11]. The type of reformulation
studied in this paper assumes that the problem QM is monotone in ξ, i.e., if x̄ is feasible for ξ̄, then
x̄ is also feasible for all ξ0 in the support of the distribution with ξ0 ≤ ξ̄. While this condition might
seem restrictive, it is actually a very general property that most applications exhibit either in their natural
formulation or after replacing some ξi j by −ξi j and changing the sign of the corresponding coefficients
si j . Note that the monotone property holds if for all j ∈ J and all i, i0 ∈ I we have si j si0 j ≥ 0, and in
particular if each random variable appears in at most one of the stochastic inequalities.
For such a monotone problem QM, an optimal solution of the proposed reformulation gives a feasible
but possibly suboptimal solution. In Section 2.2.1, we give sufficient conditions under which an optimal
solution of the reformulation gives an optimal solution of QM. These conditions are satisfied in the
applications studied in this paper, namely propagation of diseases or viruses [22, 25] and facility location
in the plane [23]. Additional examples of applications fitting the QM model are some transportation
problems [27], network flow problems with quadratic costs [63], pooling problems [42], international
supply chain problems with transfer pricing [50], siting problems of electrical substations [49], and
transmission network expansion problems [38]. This shows that the proposed approach can be used in
multiple application areas and industrial sectors.
The left-hand sides of the stochastic inequalities (3) are denoted by hi (ξ, x). They include bilinear
terms x j1 x j2 multiplied by a random variable ξi j and by a constant coefficient si j , where j = ( j1 , j2 ) ∈ J.
They are not assumed to be convex or even separable. The problem QM is general enough to encompass
the following cases:

2
(i) Some monomials involve fewer than three variables with at most one random variable and at most
two decision variables. In that case, we can add at most two artificial decision variables and at
most one random variable and fix them to 1 to fit the base formulation QM.

(ii) Some decision variables are general integer with finite upper and lower bounds instead of being
binary variables. In that case, we can replace each general integer variable by a sum of binary
variables. Note that modifying the method presented in this paper to handle directly general integer
variables is possible and is likely to be more efficient, but since the applications we consider below
involve only binary variables, we restrict the presentation to the binary case.

Probabilistic programming problems with random technology matrix have first been studied by
Kataoka [32]. He considers an individual (one-row) chance constraint with random technology matrix
where the vector ξ of random variables is assumed to be normally distributed. If the prescribed proba-
bility level p is at least 0.5, they show that the chance constraint can be equivalently reformulated as a
second-order cone one. We focus here our review on chance constraint with joint (multi-row) random
technology matrix. The complexity of these problems increases with the number |I| of inequalities that
must hold jointly. Very few results are known for constraints with multi-row random technology matrix.
They are generally limited to the case where the probabilistic constraints P(hi (ξ, x) ≥ 0, i ∈ I) ≥ p are
linear, i.e., of the form hi (ξ, x) = ξi x − di , i ∈ I. Inequalities containing products of one random variable
by one decision variable are called stochastic linear inequalities in the stochastic programming literature.
For discretely distributed random variables, Ruszczyński [56] derived cutting planes based on prece-
dence knapsack constraints and used them within a branch-and-cut algorithm. Tanner and Ntaimo [59]
proposed a Mixed-integer Linear Programming (MILP) formulation in which they incorporate irre-
ducibly infeasible optimality cuts. Beraldi and Bruni [11] and Beraldi et al. [12] have both relied upon
the concept of p-efficiency [52] and specialized branch-and-bound algorithms to solve a reformulation of
the probabilistic set covering problem. Using Monte Carlo approaches, sample approximation methods
have been shown to find good solutions for the original problem and provide statistical bounds on the
solution quality. Recently, Kogan and Lejeune [33] proposed a Boolean method to solve problems with
joint probabilistic constraints in which the elements of the multi-row random technology matrix follow
a joint probability distribution. For continuously distributed random variables, the focus has been on
the identification of conditions under which the feasible set defined by the chance constraint is convex
(see, e.g., [29, 51]) and on methods allowing for the computation of the gradients and value of multi-
variate continuous distributions, in particular the Gaussian one (see, e.g., [28, 62]). All the above papers
considered only the case where the stochastic constraints are linear, i.e., hi (ξ, x) = ξi x − di .
The method developed in this paper builds on the deterministic Boolean programming [20] reformu-
lation of linear chance constraint problems introduced in [39, 40]. The method uses for each random
variable ξi j a number of discrete values called cut points and an associated binary vector to represent
the value of ξi j . When the cut points are chosen appropriately, the mapping of the values of the random
variables in a scenario to a set of binary vectors allows for a linear description of the set of feasible
realizations. Strengths of this method are: (i) the approach can handle any type of dependency between
random variables; (ii) the method can be applied to any type (linear, quadratic, etc.) of stochastic in-
equalities; (iii) the size of the reformulated problem, in particular its number of binary variables, do not
grow linearly with the number of scenarios used to represent uncertainty, but with the number of cut
points used in the binarization process [39, 33]. Point (iii) is particularly important and sets apart the

3
binarization approach from other methods. The base reformulation can then be further simplified into a
mixed-integer (linear or not) optimization problem in which the number of binary variables or quadratic
mixed-integer terms (depending on the specifics of the problem) is only a function of the number of cut
points used in the binarization process and not on the number of scenarios. The number of cut points
grows but very slowly with the number of scenarios.
To the best of our knowledge, this study is the first one that proposes a reformulation and algorithmic
framework allowing for the solution of chance-constrained programming problem of type QM. Our
contribution span across the modeling, algorithmic, and computational fronts. It straddles stochastic,
mixed-integer nonlinear and Boolean programming as well as application areas within the Operations
Research discipline. Indeed, the models studied are relevant in multiple industrial sectors and have
extended applicability.
We present a new mixed-integer nonlinear reformulation based on Boolean modeling and derive sev-
eral sparser variants of it that provide significant computational benefits. We present detailed empirical
results comparing the various reformulations and several algorithmic ideas are implemented to provide
two new nonlinear branch-and-bound algorithms that improve the performance of the mixed-integer
nonlinear solver Couenne [9, 17]. The improvement is linked to one of the most difficult decision regard-
ing branching decisions for solving mixed-integer nonlinear programs, namely anticipating the effect of
branching on a continuous variable versus branching on a continuous one. While comparing the effect
of branching either on two integer variables or on two continuous variables can be done relatively well,
comparing between two variables of different types is empirically hard.
Guidelines on how to tune the solver and to select reformulations are presented. The test instances are
epidemiology and disaster management facility location models and cover the three types of stochastic
quadratic inequalities, namely product of two variables that are (i) both binary, (ii) binary and continuous,
or (iii) both continuous.
In Section 2, we succinctly describe the Boolean reformulation framework and propose new exten-
sion for the models studied in this paper. In Section 3, we derive a series of reformulations for the
stochastic programming problem QM. Section 4 presents the disaster management facility location and
epidemiology problems used to benchmark our method. Section 5 describes the algorithmic methods.
Section 6 details the new nonlinear branch-and-bound algorithms and analyzes the computational results,
while Section 7 provides concluding remarks.

2 Boolean Modeling Framework


In this section, we present the Boolean programming framework introduced in [33, 39, 40] that we will
extend to reformulate the type of probabilistic constraint studied in this paper. The Boolean Programming
method was initially designed to handle joint probabilistic constraints with dependent random right-hand
sides [39, 40] and later extended to stochastic programming problems with joint probabilistic constraints
and multi-row random technology matrix [33]. In [41], the Boolean framework is implemented to tackle
multiobjective probabilistic problems in which the reliability level is a decision variable. While in the
above studies all stochastic inequalities are linear, we consider here quadratic stochastic inequalities,
which poses additional challenges in terms of reformulation and solution methods. The Boolean model-
ing framework involves the construction of the set of recombinations, the binarization of the probability
distribution and the partially defined Boolean function representation of a chance constraint (Section

4
2.1), and the modeling of the feasible area of a chance constraint with a system of mixed-integer non-
linear inequalities (Section 2.2). Readers familiar with the topic will see that Section 2.1 and Section
2.2.1 are essentially based on Sections 2.1 and 2.2 in [33]. New results extending the Boolean modeling
method are presented in Section 2.2.2 and Section 3.

2.1 Recombinations, Binarization, and Boolean functions


To simplify the exposition, we assume that ξi j = ξi0 j , for all i, i0 ∈ I. The method can handle the case
where this assumption is not met without any conceptual modification, but requires heavier notation. We
thus drop the subscript i in ξi j and rewrite the chance constraint (3) as follows:
 
 X 
P  si j x j1 x j2 ξ j ≤ di , i ∈ I  ≥ p . (5)
j=( j1 , j2 )∈J

Let Ω be the set of all realizations (scenarios) ωk , k = 1, 2, . . . , |Ω| that characterize the joint probability
distribution function F of ξ.

Definition 1 [39] The realization ωk ∈ Ω is p-sufficient if and only if P(ξ ≤ ωk ) = F(ωk ) ≥ p and is
p-insufficient otherwise.

The concept of p-sufficiency defined above must not be confused with the p-efficiency concept [52] that
imposes stricter conditions. A p-sufficient realization defines sufficient conditions for (3) to hold.
Let F j be the marginal probability distribution of ξ j . The inequalities based on the univariate quantile
F j (ωkj ) ≥ p, j = 1, . . . , |J| (6)

define necessary conditions for P(ξ ≤ ωk ) ≥ p. The direct product Ω̄ = C1 × . . . × C|J| of the sets
n o
C j = ωkj : F j (ωkj ) ≥ p, k = 1, . . . , |Ω| , j ∈ J (7)

provides the set Ω̄ of recombinations [33] that defines the exhaustive list of points that can be p-sufficient
based on the univariate quantile rule. The set Ω̄ is partitioned into the set of p-sufficient recombinations
Ω̄+ := {ωk ∈ Ω̄+ : F(ωk ) ≥ p} and the set of p-insufficient recombinations Ω̄− := {ωk ∈ Ω̄− : F(ωk ) < p}.
We index the set Ω̄ (resp. Ω̄+ , Ω̄− ) with the index set K (resp. K + , K − ), i.e., Ω̄ = {ωk : k ∈ K}.
We now binarize the probability distribution and the set of recombinations with a set of values called
cut points [15]. For each j ∈ J, a collection c j1 < c j2 < . . . < c jn j of cut points are selected, with n j , j ∈ J
being the number of cut points associated with j. Then the binarization of the value ωkj maps it to the
 
vector βkj = βkj1 , . . . , βkjn j with
1 if ωkj ≥ c jl
(
β jl =
k
. (8)
0 otherwise
The binarization of a vector ωk is obtained by the binarization of each of its components. By definition,
the set of relevant Boolean vectors βk is regularized (as defined in [20]), i.e.,

βkjl ≤ βkjl0 , j ∈ J , l < l0 , k ∈ K . (9)

For the binarization to be useful, the set of cut points must be selected such that the set Ω̄+B of bina-
rizations of all recombinations in Ω̄+ is disjoint from the set Ω̄−B of those obtained from Ω̄− . A set of cut

5
points satisfying this condition is said consistent. To derive a consistent set, we use the method proposed
in [39, 40], which is itself inspired from constructive approaches stemming from the combinatorial data
mining literature [15].
Definition 2 [40] The sufficient-equivalent consistent set of cut points is given by:
|J|
[
C =
e
Cj , (10)
j=1

where the sets C j are defined by (7).


The use of the sufficient-equivalent set of cut points partitions the set of relevant Boolean vectors into the
disjoint sets of p-sufficient Ω̄+B and p-insufficient Ω̄−B relevant Boolean vectors, and permits the derivation
of a partially defined Boolean function modeling the satisfiability of the chance constraint (3).
Definition 3 [33] Given two disjoint subsets Ω̄+B , Ω̄−B : Ω̄+B Ω̄−B = Ω̄B ⊆ {0, 1}n , g Ω̄+B , Ω̄−B is a partially
S  

defined Boolean function defining the mapping (Ω̄+B Ω̄−B ) → {0, 1} such that g(k) = 1 (resp., g(k) = 0)
S

if ωk ∈ Ω̄+B (resp., Ω̄−B ).


A numerical example of application of the Boolean method is given in Appendix I.

2.2 Chance Constraint Feasibility with System of Mixed-Integer Nonlinear Inequalities


In this section, we use the partially defined Boolean function (pdBf) representation of the chance con-
straint to obtain a system of mixed-integer nonlinear inequalities enforcing the feasibility of (3). We first
list properties of threshold Boolean functions and then extend recent results obtained for the stochastic
linear case [33, 39].

2.2.1 Properties of Threshold Boolean Function

The following two definitions are useful to derive a set of mixed-integer nonlinear inequalities defining
the feasible set of the chance constraint (3).
Definition 4 [20] A function f : {0, 1}n → {0, 1} is a threshold Boolean function if, for all (a1 , . . . , an ) ∈
n
{0, 1}n , there exists λ ∈ Rn and θ ∈ R, such that f (a1 , . . . , an ) = 1 if and only if λl al ≥ θ.
P
l=1
The (n + 1)-tuple (λ, θ) is a separating structure for the threshold Boolean function f .
Definition 5 Let f be a Boolean function and let T (resp. F ) be the sets of points for which f takes
value 1 (resp. 0). The function f is said a tight minorant of a pdBf g Ω̄+B , Ω̄−B if (1) Ω̄−B ⊆ F , and (2)
T +
T Ω̄B , ∅.
Theorem 6 [33] A threshold Boolean function f defined by the separating structure (λ, θ) is a tight
minorant of a pdBf g(Ω̄+B , Ω̄−B ) if the system of inequalities
nj
for at least one k ∈ K +
XX
λ jl βkjl ≥ θ, (11)
j∈J l=1
XX nj
λ jl βkjl ≤ θ − 1, k ∈ K− (12)
j∈J l=1

has a feasible solution.

6
Inequalities (11) mean that each p-insufficient realization is also defined as p-insufficient by the separat-
ing structure (λ, θ) characterizing the tight minorant f .

Theorem 7 [33] Every feasible solution λ∗ of the system of inequalities


nj
XX
λ jl βkjl ≤ |J| − 1, k ∈ K− (13)
j∈J l=1
Xnj
λ jl = 1, j∈J (14)
l=1
λ jl ∈ {0, 1}, j ∈ J, l = 1, . . . , n j(15)

defines a threshold tight minorant f with integral separating structure (λ∗ , |J|) ∈ {0, 1}n ×Z+ for g Ω̄+B , Ω̄−B .
 

Every ωk ∈ Ω̄ such that


XX nj
λ∗jl βkjl = |J| (16)
j∈J l=1

belongs to Ω̄+ . There is at least one ωk ∈ Ω̄+ for which (16) holds.
nj
Any feasible solution λ∗ of the system (13)-(15) permits to derive a p-sufficient recombination λ∗jl c jl
P P
j∈J l=1
and can thus be used to obtain a feasible solution for (3).

2.2.2 Mixed-Integer Nonlinear Feasible Set

We now propose a new system of linear inequalities having the same properties as (13)-(15). This new
system is smaller than (13)-(15). We first construct a partially ordered set  in the space of p-insufficient
recombinations k ∈ K − .
0
Definition 8 For any ωk , ωk ∈ Ω̄− , the partial order  over the set Ω̄−B is defined by
0 0
ωk  ωk ⇔ βk ≤ βk . (17)
0
If (17) holds, the p-insufficient recombination ωk is said to be dominated by ωk . We define the set
Ω̃− of all recombinations whose binarization is dominated by at least one other (distinct) binarization of
a recombination in Ω−B . We index the set Ω̃− with K̃ − . We have the following strengthening of (13)-(15).

Lemma 9 The feasible set defined by the system of inequalities


nj
XX
λ jl βkjl ≤ |J| − 1, k ∈ K̃ − (18)
j∈J l=1

is equivalent to the one defined by (13).


0 0 nj nj 0
Proof. Consider two arbitrary ωk , ωk ∈ Ω̄−B such that ωk  ωk . Then λ jl βkjl ≤ λ jl βkjl for
P P P P
j∈J l=1 j∈J l=1
any λ jl ∈ {0, 1}. The result follows. 
+
 
Lemma 9 implies that instead of deriving a threshold tight minorant for g Ω̄B , Ω̄B , it is enough and

equivalent to derive a threshold tight minorant for g Ω̄+B , Ω̃−B .


 

7
Theorem 10 Every feasible solution of the system of inequalities {(14) − (15); (18)} defines a threshold
tight minorant with integral separating structure (λ, |J|) ∈ {0, 1} × Z+ for g Ω̄+B , Ω̃−B . Every ωk ∈ Ω̄ for
n

which (16) holds belongs to Ω̄+ .

For any pdBf g Ω̄+B , Ω̄−B , there exists a tight minorant of g Ω̄+B , Ω̄−B that is threshold [33]. An
   

identical result can be obtained and proven the same way for g Ω̄+B , Ω̃−B
 

For any pdBf g Ω̄+B , Ω̃−B , there exists a tight minorant of g Ω̄+B , Ω̃−B that is threshold.
   
Theorem 11

The next step is to use the threshold structure of the tight minorant of the pdBf g Ω̄+B , Ω̃−B to derive
 

a system of inequalities representing exactly the feasible area defined by the chance constraint (3).

Theorem 12 Any solution feasible for the set of mixed-integer trilinear inequalities

(4) ; (14) − (15) ; (18)


nj
!
λ jl c jl ≤ di , i ∈ I
P P
si j x j1 x j2 (19)
j=( j1 , j2 )∈J l=1

is feasible for the chance constraint (3).

Proof. Let ( x̂, λ̂) a feasible( solution of the above system. As ) shown in Theorem 7, λ̂ defines a threshold
nj
tight minorant f and G = ω ∈ Ω̄ : k λ̂ jl β jl = |J| ⊆ Ω̄+ .
k
P P
( j1 , j2 )∈J l=1
n o nj
Let L = ( j, l) : λ̂ jl = 1, j ∈ J, l = 1, . . . , n j . The binarization (8) implies that ωkj ≥ λ̂ jp c jp =
P
p=1
c jl , ( j, l) ∈ L, ωk ∈ G.
Furthermore, due to the definition of the sufficient-equivalent set of cut points, it is always possible
0
to find ωk ∈ G such that:
0
ωkj = c jl ( j, l) ∈ L . (20)
Since ωk ∈ G ⊆ Ω̄+B , we have P ξ ≤ ωk ≥ p, and:
0 0
 

 
 X X 0

P  si j ξ j x j1 x j2 ≤ si j ωkj x j1 x j2 , i ∈ I  ≥ p (21)
( j1 , j2 )∈J ( j1 , j2 )∈J

 
X  X 
k0
si j ω j x j1 x j2 ≤ di , i ∈ I ⇒ P  si j ξ j x j1 x j2 ≤ di , i ∈ I  ≥ p . (22)
( j1 , j2 )∈J ( j1 , j2 )∈J

Therefore, it follows from (20) and (22) that:


 nj 
X X
si j x j1 x j2  λ̂ jl c jl  ≤ di , i ∈ I

⇒ (3) holds. (23)
( j1 , j2 )∈J i=1


We now characterize in Corollary 13 a series of sufficient conditions under which the feasible sets
defined by the chance constraint (3) and the system of inequalities given in Theorem 12 are identical.

8
Corollary 13 The feasible set of the system
(4) ; (14) − (15) ; (18) − (19)

is equivalent to the feasible set of the chance constraint (3) if the set of stochastic inequalities subject to
the probabilistic requirement takes one of the following forms:

(i) Each stochastic inequality in (3) has exactly one monomial, i.e.,
 
P si j x j1 x j2 ξi ≤ di , i ∈ I ≥ p . (24)

(ii) Each stochastic inequality in (3) contains exactly one random variable, i.e.,
 
 X   
P  si j x j1 x j2 ξi ≤ di , i ∈ I  ≥ p . (25)
j=( j1 , j2 )∈J

(iii) Each stochastic inequality in (3) contains exactly one pair of decision variables xi1 xi2 :
 
 X 
P  si xi1 xi2
 ξi j ≤ di , i ∈ I  ≥ p (26)
j=( j1 , j2 )∈J

(iv) Each stochastic inequality in (3) is the product of a sum of random variables by a sum of product
of two decision variables, i.e.,
  
X  X   
P  ξi j  si j x j1 x j2 ≤ di , i ∈ I  ≥ p . (27)
j∈J j=( j1 , j2 )∈J

(v) Each stochastic inequality in (3) is a sum of trilinear terms x j1 x j2 ξi j and at most one of them can
be non-zero.

(vi) The system of inequalities subject to the probabilistic requirement includes any mix or combination
of stochastic inequalities of the form described in (i) to (v).

Proof. Theorem 12 shows that if ( x̂, λ̂) is a feasible solution of (4); (14)-(15); (18)-(19) then x̂ is
feasible for (3). We show now that there exists a p-sufficient recombination ω ∈ Ω̄ for which x̂ is feasible.
(i) Let

if si j x̂ j1 x̂ j2 = 0 if si j x̂ j1 x̂ j2 = 0
 
 ∞
  −∞

ti =  0
=
 
and t
 d i
 si j x̂ j x̂ j otherwise i  si j x̂ j x̂ j otherwise .

 di
1 2 1 2

The probability that a feasible x̂ for QM satisfies (3) is P(ξi ≤ ti , i ∈ I). Then x̂ is feasible for all ω such
that ωi ≥ ti0 . As we assume that the probability distribution is discrete, all such ω which components
ωi , i ∈ I having a positive probability of occurrence are recombinations in a set S with S ⊆ Ω̄+ . Define
ωi = min s∈S {si } for i ∈ I. Observe that ω ∈ Ω̄+ and that x̂ is feasible for ω.
(ii): Similar proof as in (i), as we have a single random variable in each stochastic inequality.
(iii): Let ζi = ξi j , i ∈ I. We can now rewrite (26) as
P
j=( j1 , j2 )∈J

P si xi1 xi2 ζi ≤ di , i ∈ I ≥ p .

(28)

9
As ξi j is discretely distributed with finite support, so is ζi . As (28) has the same form as (24), the proof
of (i) applies here too.
(iv): The chance constraint (27) can be recast as
 
 X   
si j x j1 x j2 ζi ≤ di , i ∈ I  ≥ p (29)
 0 0

P  0

j =( j1 , j2 )∈J
0 0 0 

similar to (25).
(v): For all i ∈ I, the stochastic inequality i is equivalent to the stochastic inequalities si j x j1 x j2 ≤ di for
all j ∈ J, and thus the proof of (i) applies.
(vi): The proofs of (i)-(v) essentially show that each stochastic inequality is equivalent to a collection of
inequalities of type (i). 

3 Mixed-Integer Nonlinear Programming Reformulations


In this section, we derive a series of mixed-integer nonlinear programming reformulations for problem
QM and present their characteristics.

3.1 Basic Model M1


The first reformulation M1 is obtained by a direct application of Theorem 12.

Lemma 14 The mixed-integer nonlinear programming problem M1

M1 : max qT x
subject to (2) ; (4) ; (14) − (15) ; (18) − (19)

is equivalent to the probabilistic programming problem QM.

Problem M1 is a Mixed-Integer NonLinear Programming (MINLP) problem involving trilinear terms


x j1 x j2 λ jl for j = ( j1 , j2 ) ∈ J and its continuous relaxation is in general non-convex. Such problems
are usually NP-hard to solve. Problem M1 contains n = n j binary variables λ jl as defined in (15),
P
j∈J
|Ω̃−B | knapsack constraints of form (13), |J| set partitioning constraints of form (14), and |I| nonlinear
constraints (19).
We now present a number of reformulations whose relevance depends on the properties of the trilinear
terms x j1 x j2 λ jl . These properties are here understood in terms of the continuous or binary nature of the
decision variables x j1 and x j2 . The variables λ jl are always binary. The MINLP reformulations presented
in Sections 3.2 and 3.3 are the most general ones and are always applicable regardless of whether the
variables x j1 and x j2 are continuous or not. The MILP reformulation presented in Section 3.3 is applicable
if at least one of the two variables x j1 x j2 in each trilinear term is integer.

3.2 Sparse Reformulation M2


The derivation of the model proposed in this section involves two main steps. First, we derive a new
and sparser system of mixed-integer inequalities (different from those mentioned in Section 2.2.2 and in

10
[33]) to define the feasible set of a chance constraint. Second, we use the McCormick convexification
step to reformulate the mixed-integer trilinear terms as bilinear ones.

Theorem 15 Consider an arbitrary ωk ∈ Ω̃−B and let `(


¯ j) = max{1 ≤ l ≤ n j | βk = 1}, j ∈ J. The
jl
feasible set defined by the constraints
nj
XX
λ jl βkjl ≤ |J| − 1 (30)
j∈J l=1

(14) − (15)

can be equivalently reformulated as:

γ j`(¯ j)+1 ≥ 1
P
(31)
j∈J
γ j,l−1 ≥ γ jl j ∈ J, 2 ≤ l ≤ n j (32)
γ j1 = 1 j∈J (33)
γ jl ∈ {0, 1} j ∈ J, 1 ≤ l ≤ n j (34)

using the one-to-one correspondence between λ and γ variables given by


nj
γ j,l+1 − γ j,l if l < n j
( X
λ jl = and γ jl = λ jk . (35)
γ jl if l = n j
k=l

Proof. As the transformation T given by (35) is one-to-one and linear, the polytope P corresponding
to the linear relaxation of (30); (14)-(15) is mapped to a polytope Q := T (P) and there is a one-to-one
correspondence between their respective facets. Observe that the facets of P are (30), (14), and λ jl ≥ 0
for all j ∈ J, l ∈ 1, 2, . . . , n j . Moreover, (14) is mapped to (33), and λ jl ≥ 0 for all j ∈ J, l ∈ 1, 2, . . . , n j are
mapped to (32) together with γ jn j ≥ 0. The result thus holds if (30) is mapped to (31), as the remaining
inequalities in the system 31-(34) are implied by the others.
To show that (30) is mapped to (31), observe that the constraint
nj
XX
λ jl (1 − βkjl ) ≥ 1 (36)
j∈J l=1

is equivalent to (30).
From the definition of `(
¯ j), we have 1 − β jl = 0 for l ≤ `(
¯ j), and 1 − βk = 1 for l = `(
jl
¯ j) + 1, . . . , n j .
Therefore, (36) is equivalent to
nj
X X
λ jl ≥ 1 (37)
j∈J l=`(
¯ j)+1

γ j`(¯ j)+1 ≥ 1 , proving the result.


P
Using (35) this becomes 
j∈J
Note that although the two systems in the above theorem are equivalent in term of their number of
facets, number of extreme points and integer feasible solutions, there are notable advantages to using
(31)-(34). First, the number of nonzero entries in the facet defining inequalities is smaller. This usually

11
translates in faster solution times. Second, in term of branching, one can note that branching on a variable
λ jl yields very unbalanced benefits in the two branches. In the branch where λ jl is set to 1, all variables
λ jk for k , l are set to 0 while not much progress is made in the branch where λ jl is set to 0. On the
other hand, if we branch on a variable γ jl , then in the branch where γ jl is set to 1, all variables γ jk for
k < l are set to 1 and in the branch where γ jl is set to 0, all variables γ jk for k > l are also set to 0.
The effect of the transformation is of course similar to defining the variables λ jl as an SOS1 [3] set, but
the sparsity of the system (31)-(34) has also its benefit in term of bound tightening propagation used in
mixed-integer nonlinear solvers such as Couenne. Bound tightening procedures are described in more
details in Section 5.
Pn j
In order to reformulate M1 using Theorem 15, we need to take care of the computation of l=1 λ jl c jl , j ∈
J appearing in (19). To this aim, we define the constants o jl , j ∈ J, l = 1, . . . , n j as
o jl+1 = c jl+1 − c jl , j ∈ J, l = 1, . . . , n j − 1
(
. (38)
o j1 = c j1 , j∈J
Each o jl+1 measures the distance between two consecutive cut points c jl and c jl+1 . It follows immediately
from (38) that
Xnj nj
X
o jl γ jl = c jl λ jl , j ∈ J (39)
l=1 j=1
Lemma 16 follows.
Lemma 16 Let `¯k ( j) = max{1 ≤ l ≤ n j | βkjl = 1}, j ∈ J, ωk ∈ Ω̃−B . The feasible region defined by
{(4); (14) − (15); (18); (19)} is equivalent to the one defined by the system of inequalities

γ j`¯k ( j)+1 ≥ 1, k ∈ K̃ −
P
(40)
j∈J
nj
!
γ jl o jl ≤ di , i ∈ I
P P
si j x j1 x j2 (41)
j∈J l=1
(4) ; (32) − (34) .

The mixed-integer trilinear problem M2-1

M2-1 : max qT x
subject to (2) ; (4) ; (32) − (34) ; (40) − (41)

is equivalent to M1.

We now present reformulations of M2-1 using the McCormick reformulation of bilinear terms [48] in
various ways. This assumes that the variables implied in the substitution are bounded, a typical situation.
In the application we consider, we always have 0 ≤ x j1 ≤ u j1 < ∞ for all j1 = 1, . . . , r. We thus present
the McCormick reformulation for that particular case. We linearize each mixed-integer bilinear term
x j1 γ jl in (41) and introduce an auxiliary decision variable y jl that the four McCormick inequalities

y jl ≤ u j1 γ jl j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (42)
y jl ≤ x j1 j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (43)
y jl ≥ x j1 + u j1 γ jl − u j1 j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (44)
y jl ≥ 0 j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (45)

12
force to take value x j1 γ jl . Indeed, (42) and (45) guarantee that y jl is equal to 0 if γ jl = 0, while (43) and
(44) make sure that y jl = x j1 if λ jl = 1. This, in turn, allows to replace the trilinear term x j1 x j2 γ jl by the
bilinear one x j2 y jl . The above results lead to the MINLP problem M2-1-MC presented in Lemma 17.

Lemma 17 The mixed-integer nonlinear programming problem M2-1-MC


M2-1-MC : max qT x
nj
si j x j2 y jl o jl ≤ di ,
P P
subject to i∈I (46)
j∈J l=1
(2) ; (4) ; (32) − (34) ; (40) ; (42) − (45)
is equivalent to M1.

Problem M2-1-MC is also an MINLP problem whose continuous relaxation is in general non-convex
and in which the set partitioning constraints (14) are removed and some binary variables are fixed.
In M2-1, if si j ≥ 0, we introduce the auxiliary variables e j , j ∈ J constrained by
nj
X
ej ≥ γ jl o jl , j ∈ J , (47)
l=1

in order to reduce the number of trilinear terms in the inequalities (41). Recall that x ∈ Rr+1 .
This yields the mixed-integer nonlinear programming problem M2-2
M2-2 : max qT x
si j x j1 x j2 e j ≤ di ,
P
subject to i∈I (48)
j∈J
(2) ; (4) ; (32) − (34) ; (40) ; (47)
equivalent to M1 and M2-1.

3.3 Mixed-Integer Linear Reformulations with Trilinear Terms Including at Least Two
Binary Variables
In this section, we consider the cases where the trilinear terms x j1 x j2 γ jl include at least two binary
variables. Recall that the variables γ jl are always binary. Using two McCormick reformulations, we
derive an MILP reformulation for this case. Note that the MINLP formulations derived in sections 3.1
and 3.2 remain valid.
In each trilinear term x j1 x j2 γ jl , one of the two variables x j1 or x j2 is assumed to be binary. Without
loss of generality, we assume that this variable is always x j1 . As in Section 3.2, we start with the
linearization of each bilinear term x j1 γ jl in (19) by introducing the auxiliary decision variable y jl and
the McCormick inequalities (42)-(45). As y jl is the product of two binary variables, it is automatically
binary. We can thus now apply a second round of McCormick reformulation, introducing variables v jl
for the bilinear terms y jl x j2 and the inequalities:
v jl ≤ x j2 , j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (49)
v jl ≤ y jl , j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (50)
v jl ≥ y jl + x j2 − 1 , j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j (51)
v jl ≥ 0 , j = ( j1 , j2 ) ∈ J, l = 1, . . . , n j . (52)

13
The two successive series of McCormick inequalities lead to the MILP problem M3.
Lemma 18 The mixed-integer linear programming problem M3

M3 : max qT x
subject to (2) ; (4) ; (32) − (34) ; (40) ; (42) − (52)
nj
!
v jl c jl ≤ di , i ∈ I
P P
si j (53)
j∈J l=1

is equivalent to the probabilistic programming problem QM in which for all ( j1 , j2 ) ∈ J at least one of
the two variables x j1 and x j2 is binary.

4 Application Problems
In this section we described two applications used for the computational testing of the reformulations
presented in Section 3. Algorithmic methods and specific algorithmic development for solving these
instances are described in Section 5. The problems described below are selected so that depending
on the reformulation used, we obtain a problem of type QM with trilinear terms involving either (i)
one continuous and two binary variables (Section 4.1), or (ii) two continuous and one binary variables
(Section 4.2), or (iii) three binary variables (Section 4.3).

4.1 Probabilistic Facility Location and Assignment with Random Demand: Trilinear
Terms with One Continuous and Two Binary Variables
In this section, we study a type of facility location problems in which (i) the demand is stochastic [5, 13,
16, 35]; (ii) a fixed number of facilities can be opened anywhere in the Euclidean plane [4, 10, 6]; (iii)
the distances are measured with the Manhattan (or L1 ) metric [23, 36]; (iv) the facilities are capacitated;
and (v) each customer is served by a single facility.
The objective is to minimize an upper-bound U on the weighted total-distance (i.e., the sum of the
product of the demand of each customer times the distance to the facility serving that customer) such that
the bound U is satisfied with a specified reliability level p. A related model was studied in [16] in which
the demand is normally distributed and a set of candidate locations to open the facilities is pre-defined.
Let D be the set of demand points and M be the number of facilities to be opened. The random
demand originating from d is denoted by ξd . The maximum demand level stemming from d is given by
Ud . The parameter Ci is the capacity of facility i. The binary variable yid takes value 1 if the demand
point d is served by facility i. The probabilistic total weighted distance is t =
P
td , with td denoting the
d∈D
probabilistic weighted distance to deliver to demand point d.
The facilities can be located anywhere. The Cartesian coordinates of facility i are represented by a
pair of continuous decision variables (ai , bi ), i = 1, 2, . . . , M, while the known coordinates of demand
point d are the constants (ād , b̄d ), d ∈ D. The Manhattan distance hid between facility i and demand point
d is a non-negative continuous decision variable
hid = |ai − ād | + |bi − b̄d | , i = 1, . . . , M, d ∈ D . (54)

This distance function can be linearized with the introduction of non-negative auxiliary variables sid and
s0id and using the constraints (55)-(59):

14
sid ≥ ai − ād , i ∈ I, d = 1, . . . , M (55)
sid ≥ ā j − ai , i ∈ I, d = 1, . . . , M (56)
0
sid ≥ bi − b̄d , i ∈ I, d = 1, . . . , M (57)
0
sid ≥ b̄d − bi , i ∈ I, d = 1, . . . , M (58)
0
hid = sid + sid , i ∈ I, d = 1, . . . , M (59)

The problem is thus formulated as:


P
SFL1 : min td (60)
d∈D
subject to (55) − (59)
!
M
ξd yid hid ≤ td , d ∈ D ≥ p
P
P (61)
i=1
M
yid = 1
P
d∈D (62)
i=1
M
i = 1, . . . , M
P
yid U j ≤ Ci (63)
i=1
lia ≤ ai ≤ uai i = 1, . . . , M (64)
lib ≤ bi ≤ ubi i = 1, . . . , M (65)
yid ∈ {0, 1} i = 1, . . . , M, d ∈ D (66)

The plane where the facilities can be opened is delineated which provides lower and upper bounds for the
coordinates of the facilities (see (64) and (65)). The constraints (62) stipulate that each demand point is
assigned to and served by exactly one facility. The constraints (63) ensure that the capacity of a facility
is never exceeded. Using the proposed Boolean approach, the reformulation of the above stochastic
problem takes the form of an MINLP problem and includes trilinear terms involving the product of two
binary variables by one continuous.
The instances handled in the numerical section are based on a real-life disaster network (see, e.g.,
[30, 55] that corresponds to the US Southeastern area. This region is frequently hit by hurricanes and it
is crucial to have a network enabling a timely provision of emergency commodities (water, medical kits,
food) [55]. The network includes 30 demand points (Brownsville, Corpus Christi, San Antonio, Dal-
las, Houston,Little Rock, Memphis, Biloxi, Jackson, Monroe, Lake Charles, Baton Rouge, Hammond,
New Orleans, Mobile, Birmingham, Nashville, Atlanta, Savannah, Columbia, Charlotte, Wilmington,
Charleston, Tallahassee, Lake City, Jacksonville, Orlando, Tampa, Miami, and Key West). We have gen-
erated problem instances that differ in terms of the number of scenarios (5,000, 10,000), the reliability
level enforced (0.8, 0.85, 0.9, 0.95), and the number of warehouses (2, 3, 4) that can be opened. The
AMPL and nl files corresponding to all instances used in this paper can be downloaded from [47].

4.2 Probabilistic Facility Location with Random Demand: Trilinear Terms with Two
Continuous and One Binary Variables
The difference with the model proposed in this section and the one proposed in Section 4.1 is that each
client may be served by more than one facility. The basic reformulation includes mixed-integer trilinear

15
terms including one binary variable and two continuous ones. The decision variables yi j are now defined
as continuous with lower bound 0 and upper bound 1.
The problem is formulated as:
P
SFL2 : min tj
j∈J
subject to (55) − (59) ; (61) − (65)
yi j ≥ 0 i = 1, . . . , M, j ∈ J (67)

As in Section 4.1, the test instances correspond to the real disaster network considered in [55].

4.3 Probabilistic Response Model to Propagation of Epidemics: Trilinear Terms with


Three Binary Variables
We consider here a stochastic response model that limits the propagation of epidemics or viral attacks in
a network. The recent literature devoted to this area is abundant [22, 25, 37, 54]. The Center for Disease
Control and Prevention plays an instrumental role in supporting this research and has, for example, de-
veloped resource allocation to keep under control the HIV incidence [37]. The time at which the disease
is identified, its lethality level, and the rate at which it propagates are difficult to estimate. Decisions have
to be taken under a substantial amount of uncertainty [54], and stochastic models explicitly accounting
for this uncertainty are needed.
Consider a network modeled as an undirected graph G = (V, E) with vertex set V and edge set E.
The set of nodes may correspond to locations, users, or patients, and the edges represent connections or
propagation links. Let Vc ⊆ V be a set of infected nodes and Vu = V \ Vc be the set of nodes susceptible
to be infected. The objective is to maximize the utility of the network, defined as the maximization of the
number of currently non-infected nodes that can remain open. Decisions to be taken concern the nodes
that must be shut down so that the tolerable level of threat is not exceeded. A binary decision variable
yi , i ∈ Vu defines whether node i susceptible to be contaminated can remain open (yi = 1) or must be
closed (yi = 0). The random rate at which a disease propagates from node i to v is denoted by ξi j . The
acceptable threat level as node i is upper bounded by the parameter ti , while ti j is the maximal threat
level admissible for the arc (i, j). The notation ri0 refers to the level of threat of node i if i is isolated. The
set N(i) = { j ∈ V : (i, j) ∈ E} contains the nodes directly connected to i. The problem is formulated as a
probabilistically constrained stochastic programming problem including bilinear stochastic inequalities:
P
EPID : max yi (68)
i∈Vu

ri + j∈N(i) ξi j yi y j ≤ ti
 0 P 
 
 ≥ p , i ∈ VU
 
subject to (69)

P
ξi j yi y j ≤

j ∈ N(i)

 ti j
y ∈ {0, 1}|Vu | . (70)

The reformulations M1, M2-1, and M2-2 of the above chance constraint contain trilinear terms
involving each the product of three binary variables. Each trilinear term is multiplied by a coefficient
ci jl ∈ [0, 1] which can be understood as a possible value taken by the random rate at which node i can be
infected from node j (see [25]).

16
The test instances are similar to those used in recent epidemics papers [22, 25]. The instances differ
in terms of the number of scenarios (5,000 or 10,000), the reliability level enforced (0.8, 0.85, 0.9, 0.95),
the density of the graph (10%, 20%), the maximum degree for each node in the graph (5, 10), and the
ratio of the number of infected nodes to the number of uncompromised nodes (10%, 20%, 40%).

5 Algorithmic Methods
The reformulations presented in sections 3.1 and 3.2 are nonconvex MINLP problems. Solving problems
of this type to optimality is in general difficult. Many solution techniques for MINLP have been proposed.
The surveys [8, 26, 43] give a good overview of several of these methods, and additional information can
be found in [31, 60, 61].
The method of choice for solution is the spatial Branch-and-Bound, based on convexification using
linear relaxations of the problem and applying usual Branch-and-Bound or Branch-and-Cut techniques
for solving the resulting MILP. There are several software available based on this method, commercial
ones (Baron [57], Lindoglobal [44]) or open-source (Couenne [9, 17]). The complexity of dealing with
nonlinear problems makes the performance of the solvers very unpredictable. In addition, mathematically
equivalent reformulations of an instance can also result in solution times orders of magnitude apart.
The goal of this section is to illustrate how reformulation selection and customization of a solver can
help solve problems involving chance constraints. We base our experiments on the open-source solver
Couenne, as having access to the source code allows for experiments that could not be conducted with the
other software. Due to the large number of implementation options, there is no guarantee that conclusions
drawn for Couenne are valid for another software, but it is likely that experimenting similarly with
another software would yield wide range of solution times and that default settings can be dramatically
improved when solving a very specific MINLP type of instances, such as those of interest here.

5.1 Couenne Reformulation and Bound Tightening


This section presents very briefly the framework implemented in Couenne. For more details, the inter-
ested reader is referred to [7, 9, 17]. Couenne has many options that can be activated or disabled using
an option file named couenne.opt. A file with default settings is provided with the code. All comments
in this section refer to the code in [18].
We assume that the reader is familiar with the Branch-and-bound framework for MILP, including
reliability branching [1] and strong branching [2]. The extension of Branch-and-Bound to nonlinear
programs requires two essential additional components. The first one is the construction of a linear
relaxation of the feasible set, the second one is dealing with the fact that branching on continuous vari-
ables might be needed (for example, to find the global optimum of a nonconvex problem having only
continuous variables).
To construct the linear relaxation of the problem, Couenne first builds an acyclic directed graph
representing the nonlinear expressions appearing in the instance. Each node v of the graph represents a
mathematical operation f v (from a fixed set) using all the immediate successors x1v , . . . , xtv of v. A variable
xv is associated with v and the equality xv = f v (x1v , . . . , xtv ) is associated with v. Since the possible
operations f v () are known, a linear programming convexification of the feasible set of each of them is
also known. This linearization is essentially the convex hull of feasible solutions of xv = f v (x1v , . . . , xtv )

17
and depends on the upper and lower bounds of all the involved variables. If the function f is not linear,
Couenne associates a nonlinear object to the node v. Note that imposing integrality on a variable is one
such nonlinearity and the corresponding object is called an integer object. Once all auxiliary variables xv
are introduced, we get the Couenne reformulation of the problem. Tighter bounds on the variables lead
to tighter linear relaxations, making bound tightening a crucial operation. Several options for performing
bound tightening are implemented.
Bounds on variable xk can be obtained by minimizing and maximizing the value of xk over the current
linear relaxation of the problem. This procedure is called optimality based bound tightening (OBBT).
It is the most effective bound-tightening technique, but is expensive as it involves solving 2n LPs for
a problem with n variables. The frequency at which OBBT is used is controlled by the value ` of the
parameter log num obbt per level. At a node at depth k of the enumeration tree, OBBT is applied
with probability min{1, 2`k }. The default settings for OBBT are:

optimality bt yes
log num obbt per level 1

A weaker but much faster procedure, called feasibility based bound tightening (FBBT), is also avail-
able. It amounts to bound propagation through the expression tree. The default settings for FBBT is:

feasibility bt yes

Finally, an aggressive FBBT procedure, called ABT, using information from the solution of the
nonlinear continuous relaxation for the problem is also available. The frequency at which ABT is used
is controlled by the value ` of the parameter log num abt per level in a way similar to the control of
OBBT. The default settings for ABT is:

aggressive fbbt yes


log num abt per level 2

5.2 Precision on the solution


Requirement on the precision of the solution is set to 10−6 , meaning that any linear or nonlinear expres-
sion in the Couenne reformulation must have an absolute violation smaller than that number for a solution
to be considered feasible. Ideally, one would like to find a solution feasible (within tolerances) for the
initial formulation, but as Couenne works with its reformulation, it might obtain a solution satisfying
all constraints of the reformulation within tolerances, while taking the values of the original variables in
that solution does not give a solution of the original formulation within tolerances. For example, if the
original formulation has a constraint x1 x2 + x3 x4 = 1, the Couenne reformulation is

w1 = x1 x2
w2 = x3 x4
w1 + w2 = 1

The solution x1 = x3 = 0.5, x2 = x4 = 1 + 10−6 + 10−7 , w1 = w2 = 0.5 is feasible within tolerances


for the reformulation, but this is not the case for x1 = x3 = 0.5, x2 = x4 = 1 + 10−6 + 10−7 for the
initial formulation. On the other hand, the Couenne reformulation sometimes imposes more precision

18
on the original variables than the initial formulation. For example, if x1 , x2 are binary variables while
100 ≥ x3 ≥ 0 is continuous and the initial formulation contains the equation (100x1 + x2 )x3 = 1, the
Couenne reformulation is

w1 = 100x1 + x2
w 1 x3 = 1

with the additional requirement that w1 must be integer. Adding this requirement is in general beneficial,
as it strengthen bound tightening operations and branching. However, the solution x1 = 1 + 10−6 ,
x2 = −10−6 , x3 = 10−2 is within tolerances for the initial formulation while 100x1 +x2 = 100+10−4 −10−6
is not within precision to be considered integer.
Couenne of course checks if any solution it runs into is feasible within tolerance for the initial for-
mulation. Couenne updates the upper bound when this is the case, but it has still to process the node
if its lower bound does not match the upper bound. For most problems, this has little practical impact,
the difference in objective value of slightly infeasible solutions being close to each other. However, to
avoid having Couenne branch sometimes very deep to obtain a matching upper and lower bound we set a
tolerance for the final gap between upper and lower bounds to 0.5% of the upper bound using the option

bonmin.allowable fraction gap 0.005

5.3 Scaling of constraints and objective function


As the precision of the solution is set to 10−6 , and Couenne does not rescale the constraints or variables,
the scaling of the original constraints is very important. For the facility location problem instances,
the demand for commodities varies a lot between locations (i.e., between 10 and 15000 units), and its
magnitude differs from the values taken by the decision variables representing the coordinates of the
opened facilities and the distances between demand points and facilities. We have rescaled the demand
so that it takes value in (0, 100]. The problem instances corresponding to the epidemic propagation model
4.3 did not need any scaling.

5.4 Strong Branching vs. Osi Branching


Deciding on which variable to branch is a crucial and difficult decision, more so in spatial Branch-and-
Bound than in “regular” Branch-and-Bound where usually branching variables are selected among the
integer ones. At a given node of the Branch-and-Bound, Couenne computes the optimal solution x̄ to the
linear relaxation at the node, scans the nonlinear objects, and flags the violated ones (i.e., those for which
x̄v , f v ( x̄1v , x̄2v , . . . , x̄tv ). It then has two major options to select the branching variable:

• Osi branching. With this option, Couenne essentially selects a most violated nonlinear object,
applies rules to select one variable involved in the object, and chooses a way to split the current
domain of the variable in two parts. This is very fast, but usually provides a poor choice of
branching variable. On easy problems, this option might be the fastest, but on hard problems this
is rarely the case.

• Strong branching. As for MILP, strong branching in Couenne works by building a list of candidate
branchings, compute the lower bounds resulting of each of the branchings and then select the most

19
promising one according to a scoring function. The candidate selection is quite involved. It may
use essentially two methods:

– Reliability Branching. Introduced by Achterberg et al. [1] in the context of MILP, it can
be generalized to MINLP, but this generalization is far from unique. It relies on the idea
of pseudo-cost, but computation of reliable pseudo-cost is very challenging for MINLP.
Couenne implementation of pseudo-cost is unsatisfactory and we disable it, turning relia-
bility branching into strong branching.
– Violation Transfer. Proposed by Tawarmalani and Sahinidis [61], its implementation in
Couenne is a variant of the original version, due to internal constraints on the branching
selection process in Couenne. The basic idea is to scan the violated objects and attribute part
of the violation of the object to each of the variables involved in the corresponding expres-
sion. The reader might consult [9, 61] for details.

The choice between Osi and Strong branching is made using one of the two options

variable selection osi-simple


variable selection osi-strong

With the latter option, candidates can be selected using reliability branching or violation transfer using
respectively one of the options

branching object var obj


branching object vt obj

If strong branching is used, the parameters for selecting the number of candidates to consider at the root
(say, 30) and at other nodes of the enumeration tree (say, 20) are set using the options

bonmin.number strong branch root 30


bonmin.number strong branch 20

Disabling the use of pseudo-costs for all practical purposes can be done using

bonmin.number before trust 1000000

5.5 Branching First on Classes of Variables


When using strong branching, the candidate selection is a rule involving a mix of a priority order on
the nonlinear objects (integer objects being preferred), and a usefulness criterion, trying to estimate the
impact of branching on the object. It turns out that, for some classes of problems, selecting as candidates
for strong branching only violated integer objects when some are violated, and using other nonlinear
objects only when all integer objects are satisfied is much better than the default candidate selection.
While it is possible to select different priority for continuous and integer variables in the couenne.opt
file, this does not guarantee that branching will occur on an integer variable when one violated integer
object exists. It only guarantees that the strong branching candidate list will be first filled with violated

20
integer objects and if space in the list is still available, then additional variables involved in violated non-
integer object might be selected. We implemented the requirement that branching on an integer variable
with fractional value in the current LP solution must occur if one exist by modifying the strong branching
candidate selection. We call this modification Branching First on Integer (BFI).
Branching candidates selected by Couenne are always among variables appearing in a nonlinear
object. The justification for this is that all linear constraints of the problem are in the linear relaxation
used by Couenne, so branching to make sure that all nonlinear expressions are satisfied is sufficient for
correctness. But in specific situations, branching on variables that are not involved in any nonlinear
expression is far superior (see in Section 6.1). Modifications to the code of Couenne to implement this
option are non trivial. We coded this option by modifying the strong branching candidate selection and
the computations of violations for nonlinear objects and call it Branching for Facility Location (BFL).

5.6 Tight Variable Bounds


Variables with no lower or no upper bounds create numerical problems for bound propagation. As
much as possible, giving finite bounds on all variables is advisable and can make a huge performance
difference. For each instance considered in this paper, we have computed the tightest lower and upper
bounds that can be assigned a priori to each decision variable and have explicitly listed these bounds in
the initial formulation provided to Couenne.

6 Computational Results
In this section, we evaluate the empirical strength of the proposed reformulation and algorithmic methods
for the applications described in Section 4.
The binarization process employed for deriving the proposed MILP formulations is implemented
in C++. The AMPL modeling language is used to formulate the mathematical programming prob-
lems. The mixed-integer programming formulations are solved with the Cplex 12.6 solver and the
mixed-integer nonlinear programming ones (with nonconvex continuous relaxations) are solved with
the Couenne solver. The machine used is a 64 bits PowerEdge R515 with twelve AMD Opteron 4176
2.4GHz processors, of which we used only one, with 64GB of memory, the Linux Fedora 19 operating
system, and gcc version 4.8.2 20131212 (Red Hat 4.8.2-7) compiler.

6.1 Stochastic Facility Location Model – Part I: Trilinear Terms with Two Continuous
and One Binary Variables
This section describes the steps typically needed when trying to evaluate the strength of various MINLP
formulations and algorithmic options using the Couenne solver. We start with the default solvers dubbed
St and Tr (out-of-the-box codes from [19] (stable version) and [18] (trunk version)) and continues with
two new variants Trb and Trb-I of the latter, altering the strong branching code in both and in addition
with Trb-I using branching first on integer variables (BFI) as described in Section 5.5. All solvers use
strong branching for branching decisions. Unless otherwise specified, the parameters for maximum cpu
time, branching object type, relative optimal gap, strong branching, and bound tightening are set as
follows (see Section 5.1 for the explanation of the parameters). All other parameters are set to their
default value:

21
• time limit 9000

• bonmin.allowable fraction gap 0.005

• branching object vt obj

• bonmin.number strong branch root 20

• bonmin.number strong branch 2

• bonmin.number before trust 1000000

• feasibility bt yes

• optimality bt no

6.1.1 Preliminary Results with Defaults Solvers

For the Facility Location Problem with possible split satisfaction of demands described in Section 4.2,
the basic formulation is M1 containing trilinear terms that are products of a binary variable (γ j` , from
the discretization of the stochastic variable) with two continuous variables (hid , for the distance between
facility i and demand point d, and yid , for the fraction of the demand of customer d satisfied by facility i).
As discussed in Section 3.2, the reformulations M2-2 are expected to work better than the reformulation
M1. However, results obtained by the four codes St, Tr, Trb, and Trb-I on M1 and M2-2 are horrible,
none of the codes solving any of the instances within the time limit of 9,000 seconds.

6.1.2 Nonlinear Branch-and-Bound Algorithms Trb-L and Trb-L-I

When studying the branching decisions made during the tests with Trb and Trb-I, we noticed that the
codes branch often on variables hid or yid . Due to bound propagation techniques used by Couenne and
the presence of equations (54), it should be much better to branch on variables ai or bi instead of hid .
Indeed, ai and bi are involved in equation (54) for all customers and reducing the range of one of them
will, in general, reduce the ranges of all variables hid for all d. On the other hand, branching on variable
hid for one particular d might reduce or not the range for variable ai or bi and is likely to induce far less
bound tightening. This option is called Branching for Facility Location (BFL).
A second improvement is generated by the following simple observation. Let R be the ranges of all
variables hid at a node of the enumeration tree. Let G(R) be the bipartite graph whose vertices corresponds
to the facilities and demand points and containing an edge between facility i and demand point d if and
only if the range of variable hid excludes both 0 and 1, i.e. if and only if hid must take a fractional value
in any feasible solution. Then there exists an optimal solution with ranges S such that G(S ) is acyclic.
Thus, any node of the enumeration tree for which G(R) is not acyclic can be pruned.
Based on these two observations, we modified the codes of Trb and Trb-I such that when Couenne
wishes to include a variable hid in its strong branching candidates, it includes either ai or bi instead (the
one with largest range) and prune nodes according to the second observation. This gives us two new
codes specifically designed to solve these facility location instances named Trb-L and Trb-L-I. Note
that Trb-L-I applies first BFL and if no variable is selected, then it uses BFI, and then regular candidate
selection.

22
6.1.3 Results with Trb-L and Trb-L-I

We characterize each problem instance with the notation bB cC M p Ω, where b is the number of binary
variables in the trilinear terms, c is the number of continuous variables in the trilinear terms, M is the
number of facilities to be opened, p is the enforced reliability level, and Ω is the number of scenarios
used to represent the joint distribution of the random variables.

Table 1: Results for Trb-L and Trb-L-I for formulations M1, M2-1, and M2-2.
Trb-L Trb-L-I
M1 M2-1 M2-2 M1 M2-1 M2-2
1B2C 2 95 5000 724.3 11.9 8.7 1825.0 13.4 12.6
1B2C 3 95 5000 507.3 353.1 331.3 335.6 621.5 797.2
1B2C 4 95 5000 9000 4700.3 3868.3 9000 3803.6 4302.1
1B2C 2 90 5000 9000 119.7 93.2 9000 254.2 212.1
1B2C 3 90 5000 9000 3743.5 2841.4 9000 9000 9000
1B2C 4 90 5000 9000 9000 9000 9000 9000 9000
1B2C 2 85 5000 9000 509.1 587.3 9000 532.2 571.7
1B2C 3 85 5000 9000 9000 9000 9000 9000 9000
1B2C 4 85 5000 9000 9000 9000 9000 9000 9000
1B2C 2 80 5000 9000 9000 4820.5 9000 3192.7 3393.4
1B2C 3 80 5000 9000 9000 9000 9000 9000 9000
1B2C 4 80 5000 9000 9000 9000 9000 9000 9000
1B2C 2 95 10000 422.2 15.2 14.0 524.5 21.8 17.5
1B2C 3 95 10000 9000 477.7 401.2 9000 492.0 488.9
1B2C 4 95 10000 9000 3585.7 3588.3 9000 4075.6 9000
1B2C 2 90 10000 9000 141.5 127.9 9000 187.2 253.4
1B2C 3 90 10000 9000 3356.7 3342.0 9000 4608.7 6936.6
1B2C 4 90 10000 9000 9000 9000 9000 9000 9000
1B2C 2 85 10000 9000 706.1 648.0 9000 771.1 1023.9
1B2C 3 85 10000 9000 9000 9000 9000 9000 9000
1B2C 4 85 10000 9000 9000 9000 9000 9000 9000
1B2C 2 80 10000 9000 9000 9000 9000 5176.6 4348.7
1B2C 3 80 10000 9000 9000 9000 9000 9000 9000
1B2C 4 80 10000 9000 9000 9000 9000 9000 9000
Average Time 7943.9 5238.4 4986.3 7986.9 5114.6 5431.6
Minimal Time 507.3 11.9 8.7 335.6 13.4 12.6
Maximal Time 9000 9000 9000 9000 9000 9000
Unsolved Instances 21 12 11 21 11 12

Table 1 clearly shows that both formulations M2-1 and M2-2 are superior to M1. Only 4 in-
stances can be solved in three hours with the M1 formulation. Solving M2-2 with Trb-L is clearly
the formulation-algorithm combination that works best.
We evaluate the difficulty of solving a problem instances in terms of the computational time and the
number of instances that could not be solved to optimality within 3 hours. The cpu time and number of
unsolved instances increase with the number M of facilities that are opened and as the probability level p
decreases. This is expected as this is directly related to the number of trilinear terms in the reformulation.

23
On the other hand, note that the cpu time and number of unsolved instances is not much impacted by the
number of scenarios used to represent uncertainty. The first 12 instances in Table 1 and the last 12 can be
paired in the obvious way such that the only difference in a pair is the number of scenarios used, 5,000
vs. 10,000. If both instances in a pair can be solved within the time limit, the instance using 10,000
scenarios requires about 25% more cpu time than its mate, far from the typical super-linear increase for
other methods (see, e.g., the method proposed in [34] for linear stochastic inequalities with random right-
hand sides and deterministic technology matrix). This is a major advantage of the proposed method. This
reflects the fact the employed Boolean method employed provides reformulations in which the number
of included binary variables does not depend directly on the number of scenarios. This further translates
into the fact that the number of trilinear terms involving these binary variables is not a function of the
number of scenarios.
When formulation M2-2 is passed to Couenne, it breaks the trilinear term hid · yid · γd` by creating a
new variable for vid = hid · yid and then a variable zid` = vid · γd` . Another possibility is to introduce in
M2-2 a new variable vid` = γd` · yid and apply a McCormick reformulation on these equalities and then
a variable zid` = vid · hid . This gives us formulation M2-2-MC1. A third possibility is to introduce in
M2-2 a new variable vid` = γd` · hid and apply a McCormick reformulation on these equalities and then
a variable zid` = vid · yid . This gives us formulation M2-2-MC2.
Table 2 gives the results obtained by Trb-L on the formulations M2-2, M2-2-MC1, and M2-2-MC2.
It is worth noting that Trb-L-I does not perform well on the M2-2-MC1, and M2-2-MC2 formulations,
being unable to solve any of the instances. Performances on M2-2 are slightly better than with M2-2-
MC2, and much better than with M2-2-MC1.
At first sight, it is surprising to observe such a difference between the reformulations M2-2-MC1
and M2-2-MC2. Both involve the incorporation of the same number of auxiliary variables due to the
linearization of bilinear terms with the McCormick approach. One likely reason for the difference is
that since Trb-L looks first at violated nonlinear objects involving variables hid , in M2-2-MC1 it might
branch on the objects zid` = vid` · hid while the variables vid` are not set to values close to satisfy vid` =
γd` · yid . This creates useless branches. On the other hand, in M2-2-MC2, Trb-L branches more logically
first on the objects vid` = γd` · hid and only when these objects are satisfied, it moves on to branching on
the objects zid` = vid · yid . This type of ordering information between objects is available in Couenne, but
is not used in branching decisions. This observation might lead to improvement in branching decisions
in nonlinear solvers as it has relevance beyond the very specific instances at hand.
Next, we investigate the effect of changing some of the parameters of Trb-L. First, we look at three
settings for the following parameters:

• bonmin.number strong branch root: 30, 20 or 10

• bonmin.number strong branch: 10, 5, or 2.

This gives us 9 parameter settings labeled (a, b) where a (resp. b) is the value for the first (resp.
second) parameter. The results are in Table 8 in the appendix. The variation in the results are not
dramatic, with a difference of about 10% in computing time between best and worst, but the number
of instances not solved (8) is smallest for the settings of (10, 10), the other variants having a number of
unsolved instances ranging between 9 and 11.
Finally, we look at variations of the bound tightening options of Couenne, namely

24
Table 2: Results for Trb-L on formulation obtained from M2-2 by various McCormick reformulations.
M2-2 M2-2-MC1 M2-2-MC2
1B2C 2 95 5000 8.7 1962.3 15.6
1B2C 3 95 5000 331.3 9000 806.2
1B2C 4 95 5000 3868.3 9000 9000
1B2C 2 90 5000 93.2 9000 140.7
1B2C 3 90 5000 2841.4 9000 4646.0
1B2C 4 90 5000 9000 9000 9000
1B2C 2 85 5000 587.3 9000 1008.4
1B2C 3 85 5000 9000 9000 9000
1B2C 4 85 5000 9000 9000 9000
1B2C 2 80 5000 4820.5 9000 3080.7
1B2C 3 80 5000 9000 9000 9000
1B2C 4 80 5000 9000 9000 9000
1B2C 2 95 10000 14.0 9000 16.3
1B2C 3 95 10000 401.2 9000 595.1
1B2C 4 95 10000 3588.3 1893.2 7347.2
1B2C 2 90 10000 127.9 9000 69.1
1B2C 3 90 10000 3342.0 9000 4826.0
1B2C 4 90 10000 9000 9000 9000
1B2C 2 85 10000 648.0 9000 624.2
1B2C 3 85 10000 9000 9000 9000
1B2C 4 85 10000 9000 9000 9000
1B2C 2 80 10000 9000 9000 5630.2
1B2C 3 80 10000 9000 9000 9000
1B2C 4 80 10000 9000 9000 9000
Average Time 4986.3 8410.6 5325.2
Minimal Time 8.7 1893.2 15.6
Maximal Time 9000 9000 9000
Unsolved Instances 11 22 11

• feasibility bt: yes or no

• optimality bt: yes or no.

This gives us 4 parameter settings, labeled “yy”, “yn”, “ny”, and “nn”, the first letter referring to the
setting of feasibility bt. The results are similar to those for varying the strong branching parameters
(see Table 9 in the appendix), with the setting “yn” being the fastest by a small margin but solving 16 out
of 24 instances within the time limit, all other variants solving fewer.

6.2 Stochastic Facility Location Model – Part II: Trilinear Terms with Two Binary and
One Continuous Variables
When the demand of a customer is required to be satisfied from a single facility (see Section 4.1), we
obtain a problem with trilinear terms identical to those in the previous section, but each term has now
two binary and one continuous variables. We start with testing the four codes described in the previous

25
section, together with the Stable and Tr codes on the formulation M2-2, that proved to be vastly superior
to M1 in the previous section.

Table 3: Results for codes with formulation M2-2.


Stable Tr Trb-I Trb Trb-L-I Trb-L
2B1C 2 95 5000 20.5 11.2 4.5 22.6 6.1 20.7
2B1C 3 95 5000 110.3 244.9 43.4 197.2 39.1 2489.9
2B1C 4 95 5000 446.8 9000 106.2 644.5 72.6 9000
2B1C 2 90 5000 15435.2 57.99 48.2 50.9 17.8 9000
2B1C 3 90 5000 1183.1 852.62 65.2 535.4 65.8 9000
2B1C 4 90 5000 9000 9000 888.1 9000 197.7 9000
2B1C 2 85 5000 9000 142.0 176.1 254.4 43.2 9000
2B1C 3 85 5000 9000 9000.3 1091.0 1512.7 425.6 9000
2B1C 4 85 5000 9000 9000 9000 9000 3621.5 9000
2B1C 2 80 5000 9000 1107.8 1226.9 2155.4 304.0 9000
2B1C 3 80 5000 9000 9000 7017.6 8070.9 4521.5 9000
2B1C 4 80 5000 9000 9000 9000 9000 9000 9000
2B1C 2 95 10000 25.2 12.2 8.8 18.2 7.1 38.4
2B1C 3 95 10000 148.4 543.4 41.1 188.4 39.4 9000
2B1C 4 95 10000 450.2 9000 103.9 472.0 82.5 9000
2B1C 2 90 10000 147.2 55.0 54.3 138.8 18.6 9000
2B1C 3 90 10000 1515.5 1102.8 116.2 710.9 122.9 9000
2B1C 4 90 10000 9000 4010.8 843.7 3033.8 375.0 9000
2B1C 2 85 10000 1080.3 350.1 59.9 491.4 78.6 9000
2B1C 3 85 10000 9000 2907.0 2055.1 2009.0 549.3 9000
2B1C 4 85 10000 9000 9000 9000 9000 9000 9000
2B1C 2 80 10000 9000 9000 1119.5 9000 387.5 9000
2B1C 3 80 10000 9000 9000 9000 9000 4721.3 9000
2B1C 4 80 10000 9000 9000 9000 9000 9000 9000
Average Time 5731.8 4599.9 2502.9 3479.4 1779.1 7981.2
Minimal Time 20.5 11.2 4.5 18.2 6.1 20.7
Maximal Time 15435 9000 9000 9000 9000 9000
Unsolved Instances 14 11 5 7 3 21

Table 3 shows that the codes branching first on integer variables (Trb-I and Trb-L-I) dominates the
other, and between these two, the latter is the best on formulation M2-2. We nevertheless present tests for
both codes on alternative formulations, due to the difficulty of estimating the importance of branching
first on integer variables, as Trb-I does, or first on facility coordinates and then integer variables as
Trb-L-I does.
We now report the results for three reformulations of M2-2 obtained by using McCormick lineariza-
tions of various bilinear terms:

• formulation M2-2-MC0 obtained by introducing in M2-2 a new variable vid = hid ·yid and applying
the McCormick reformulation on these equalities.

• formulation M2-2-MC1 obtained by introducing in M2-2 a new binary variable vid` = γd` · yid and

26
applying the McCormick reformulation on these equalities.

• formulation M2-2-MC2 obtained by introducing in M2-2 a new variable vid` = γd` · hid and
applying the McCormick reformulation on these equalities.

The last two reformulations are identical to those presented in the previous section except that here
variables y are binary, justifying that we re-use these names.

Table 4: Results for Trb-L-I on various McCormick reformulations.


M2-2 M2-2-MC1-0 M2-2-MC1-1 M2-2-MC1-2
2B1C 2 95 5000 6.1 4.2 12.3 4.3
2B1C 3 95 5000 39.1 34.1 9000 35.8
2B1C 4 95 5000 72.6 354.4 258.2 96.5
2B1C 2 90 5000 17.8 11.6 56.8 10.2
2B1C 3 90 5000 65.8 68.4 530.3 90.3
2B1C 4 90 5000 197.7 795.9 1669.6 185.8
2B1C 2 85 5000 43.2 25.5 170.9 34.1
2B1C 3 85 5000 425.6 359.8 9000 248.7
2B1C 4 85 5000 3621.5 1023.2 816.3 761.0
2B1C 2 80 5000 304.0 61.9 9000 152.5
2B1C 3 80 5000 4521.5 1578.0 9000 2092.1
2B1C 4 80 5000 9000 1982.3 9000 2916.5
2B1C 2 95 10000 7.1 4.7 12.6 4.2
2B1C 3 95 10000 39.4 49.9 187.0 610.6
2B1C 4 95 10000 82.5 140.4 592.3 122.6
2B1C 2 90 10000 18.6 11.8 36.2 16.2
2B1C 3 90 10000 122.9 62.9 451.8 1472.3
2B1C 4 90 10000 375.0 368.0 9000 940.2
2B1C 2 85 10000 78.6 29.3 200.6 42.3
2B1C 3 85 10000 549.3 307.6 9000 726.5
2B1C 4 85 10000 9000 581.6 9000 463.1
2B1C 2 80 10000 387.5 134.7 1761.3 129.2
2B1C 3 80 10000 4721.3 6233.4 9000 5088.7
2B1C 4 80 10000 9000 1542.7 9000 6395.7
Average Time 1778.7 656.9 4030.7 943.3
Minimal Time 6.1 4.2 12.3 4.2
Maximal Time 9000 6233.4 9000 6395.7
Unsolved Instances 3 0 10 0

Table 4 shows that the reformulation M2-2-MC0 improves results of Trb-L-I significantly. Note that
the discussion on performances of Trb-L-I on M2-2-MC1 and M2-2-MC2 in the previous section seems
to be pertinent here: M2-2-MC0 and M2-2-MC2 are expected to perform better than M2-2-MC1 due
to prioritization of branching. In additions, M2-2-MC0 introduces fewer nonlinear objects and this is
probably the explanation for its better performances.
Table 5 shows that the reformulation M2-2-MC0 improves results of Trb-I even more, so that on this
formulation Trb-I is clearly the best performer. It is difficult to explain why the performances of Trb-I

27
Table 5: Results for Trb-I on various McCormick reformulations.
M2-2 M2-2-MC0 M2-2-MC1 M2-2-MC2
2B1C 2 95 5000 4.5 5.0 13.8 3.9
2B1C 3 95 5000 43.4 29.8 170.8 28.5
2B1C 4 95 5000 106.2 418.6 267.4 228.6
2B1C 2 90 5000 48.2 10.1 53.9 12.2
2B1C 3 90 5000 65.2 77.3 716.2 90.7
2B1C 4 90 5000 888.1 384.4 3047.7 217.7
2B1C 2 85 5000 176.1 28.5 565.3 32.6
2B1C 3 85 5000 1091.0 388.3 9000 342.4
2B1C 4 85 5000 9000 835.5 9000 1158.9
2B1C 2 80 5000 1226.9 56.7 1775.7 132.1
2B1C 3 80 5000 7017.6 1239.8 9000 892.4
2B1C 4 80 5000 9000 1617.0 9000 2229.7
2B1C 2 95 10000 8.8 5.1 11.7 4.0
2B1C 3 95 10000 41.1 48.1 156.8 606.4
2B1C 4 95 10000 103.9 160.5 567.9 145.8
2B1C 2 90 10000 54.3 10.0 54.3 17.6
2B1C 3 90 10000 116.2 60.9 659.6 212.5
2B1C 4 90 10000 843.7 371.0 2950.3 317.2
2B1C 2 85 10000 59.9 29.5 363.8 37.7
2B1C 3 85 10000 2055.1 329.3 9000 590.4
2B1C 4 85 10000 9000 567.3 9000 498.1
2B1C 2 80 10000 1119.5 110.9 4177.3 128.9
2B1C 3 80 10000 9000 1507.3 9000 6039.1
2B1C 4 80 10000 9000 1609.6 9000 3561.1
Average Time 2502.9 412.5 3646.7 730.4
Minimal Time 4.5 5.0 11.7 3.9
Maximal Time 9000 1617 9000 6039.1
Unsolved Instances 5 0 8 0

on M2-2-MC1-1 are so much worse than on the other two formulations. The McCormick reformulation
used to obtain M2-2-MC1-1 gives the convex hull of the feasible points for the three binary variables
involved. One might expect that this would lead to a easier problem to solve, as the nonlinear constraints
become quadratic instead of trilinear, but this is not the case.
Next, we investigate the effect of changing two of the strong branching parameters as explained in
the previous section. For both Trb-I and Trb-L-I the settings (10, 2) are best. For average cpu time,
difference between best and worst setting is a factor of up to 3, with the best setting solving all instances
(see Table 10 and 11 in the appendix for full details.)
Finally, we look at variations of the bound tightening options feasibility bt and optimality bt,
as described in the previous section. For Trb-L-I, the best options are “yn” by a small margin. For Trb-I
all settings are more or less equivalent (see Table 12 and 13 in the appendix for full details).
As aforementioned, we can apply two rounds of McCormick linearizations if the trilinear terms
include at least two binary variables. We can then obtain the MIP linear formulation M3 that can be

28
solved with Cplex. Note that Cplex is about eight times faster than Couenne on average.

6.3 Stochastic Epidemics Propagation Model: Trilinear Terms with Products of Three
Binary Variables
For the epidemics problem instances, we only report the results obtained by solving the MIP linear
reformulation obtained by applying two rounds of McCormick linearizations with the Cplex solver. All
instances are solved in about 8 seconds on average. This is much faster than anything that can be obtained
using Couenne. In general, using an MILP reformulation and Cplex is likely to be the fastest way to solve
instances of this type.

7 Conclusion
This paper present reformulations, algorithms, and empirical guidelines for solving a complex class of
probabilistically constrained stochastic programming problems QM. To our knowledge, it is the first
study that proposes a method to solve that class of stochastic nonlinear programming problems. This
class covers applications in multiple areas ranging from epidemiology [22, 25], transportation [27], sup-
ply chain [50], natural gas and refinery [42], or energy [38], to name a few. The proposed reformulation
method is particularly appealing, since: (i) it can tackle any type of dependency between random vari-
ables; (ii) it is applicable in a uniform way if the stochastic inequalities are linear, quadratic, or include
polynomials of higher degree; (iii) the size of the reformulation, and in particular the number of binary
variables and nonlinear terms do not increase linearly with the number of scenarios used to represent
uncertainty.
The theoretical contributions of this study span across the stochastic, mixed-integer nonlinear and
Boolean programming disciplines. We propose a Boolean programming approach to derive determin-
istic reformulations for problems in QM with trilinear inequalities with terms involving the product of
two decision variables with one random variable. We classify problems in QM according to the types
(continuous or binary) of decision variables appearing in the trilinear terms.
We evaluate computationally the performance of the proposed reformulations on facility location
and epidemic propagation problems, covering the three main problem classes in QM. For problems with
trilinear terms involving two continuous decision variables, we modify the behavior of the generic solver
Couenne and improve its performances dramatically. We show how the knowledge of the inner-working
of a software can give clues on how to improve its performances on a particular class of problems.
We also derive potential explanations for the superiority of some reformulations over others. These
observations, if valid in a larger context, could help in developing rules-of-thumbs for selecting reformu-
lations for MINLP problems. We also evaluate several priority order for selecting branching variables
and observe that matching the formulation used and the branching rule is important. When a problem in
QM has trilinear terms that involve at least one binary decision variable, our advice is to use two rounds
of McCormick linearization to obtain an MILP problem that can be solved with an MILP solver such as
Cplex. If both decision variables appearing in the trilinear terms are continuous, we observe that the best
option is problem dependent. This should serve as a reminder that a fair empirical comparison between
a tailored made code with a generic solver such as Couenne on a class of instances is difficult to obtain,
as minor modifications of the generic code might improve its performances dramatically.

29
References
[1] Achterberg T., Koch T., Martin A. 2005. Branching Rules Revisited. OR Letters 33, 42–54.
[2] Applegate D.L., Bixby R.E., Chvátal V., Cook W.J. 2006. The Traveling Salesman Problem, A
Computational Study. Princeton.
[3] Beale E. M. L., Forrest J. J. H. 1976. Global Optimization Using Special Ordered Sets. Mathemat-
ical Programming 10, 52–69.
[4] Averbakh I., Bereg S. 2005. Facility Location Problems with Uncertainty on the Plane. Discrete
Optimization 2 (1), 3–34.
[5] Baron O., Berman O., Krass D. 2008. Facility Location with Stochastic Demand and Constraints
on Waiting Time. Manufacturing & Service Operations Management 100 (3), 484–505.
[6] Baron O., Berman O., Krass D., Wang Q. 2007. The Equitable Location Problem on the Plane.
European Journal of Operational Research 90, 183–578.
[7] Belotti P. 2009. Couenne: A User’s Manual. Technical Report, Lehigh University.
[8] Belotti P., Kirches C., Leyffer S., Linderoth J., Luedtke J., Mahajan A. 2013. Mixed-Integer Non-
linear Optimization. Acta Numerica 22, 1–131.
[9] Belotti P., Lee J., Liberti L., Margot, F., Wächter, A. 2009. Branching and Bound Tightening Tech-
niques for Nonconvex MINLPs. Optimization Methods and Software 24, 597–634.
[10] Benkoczi R., Bhattacharya B.K., Das S., Sember J. 2009. Single Facility Collection Depots Loca-
tion Problem in the Plane. Computational Geometry 42 (5), 403–418.
[11] Beraldi P., Bruni M.E. 2010. An Exact Approach for Solving Integer Problems Under Probabilistic
Constraints with Random Technology Matrix. Annals of Operations Research 177 (1), 127–137.
[12] Beraldi P., Bruni M.E., Violi A. 2012. Capital Rationing Problems Under Uncertainty and Risk.
Computational Optimization and Applications 51 (3), 1375–1396.
[13] Berman O., Krass D. 2011. On n-Facility Median Problem with Facilities Subject to Failure Facing
Uniform Demand. Discrete Applied Mathematics 159, 420–432.
[14] Bonami P., Lejeune M.A. 2009. An Exact Solution Approach for Portfolio Optimization Problems
Under Stochastic and Integer Constraints. Operations Research 57 (3), 650-670.
[15] Boros E., Hammer P.L., Ibaraki T., Kogan A. 1997. Logical Analysis of Numerical Data. Mathe-
matical Programming 79 (1-3), 163–190.
[16] Carbone R. 1974 Public Facilities Location under Stochastic Demand. INFOR 12 (3), 261–270.
[17] COIN-OR. Computational Infrastructure for Operations Research Project.
http://www.coin-or.org/
[18] https://projects.coin-or.org/svn/Couenne/trunk, revision 1070.
[19] https://projects.coin-or.org/svn/Couenne/stable/0.4, revision 1068.
[20] Crama Y., Hammer P.L. 2011. Boolean Functions - Theory, Algorithms, and Applications. Cam-
bridge Press University.
[21] Dentcheva D., Prékopa A., Ruszczyński A. 2001. Concavity and Efficient Points of Discrete
Distributions in Probabilistic Programming. Mathematical Programming 47 (3), 1997–2009.
[22] Enns E.A., Mounzer J.J., Brandeau M.L. 2012. Optimal Link Removal for Epidemic Mitigation: A
Two-Way Partitioning Approach. Mathematical Biosciences 235, 138-147.

30
[23] Farahani R.Z., Hekmatfar M. 2009. Facility Location: Concepts, Models, Algorithms and Case
Studies. Springer, Heidelberg, Germany.
[24] Geletu A., Kloeppel M., Zhang H., Li P. 2013. Advances and Applications of Chance-Constrained
Approaches to Systems Optimization under Uncertainty. International Journal of Systems Science
44, 1209–1232.
[25] Goldberg N, Leyffer S., Safro I. 2012. Optimal Response to Epidemics and Cyber Attacks in
Networks. Preprint Mathematics and Computer Science Division, Argonne National Laboratory
ANL/MCS-1992-0112, 1–24.
[26] Grossmann I. 2002. Review of Nonlinear Mixed-Integer and Disjunctive Programming Techniques.
Optimization and Engineering 3, 227–252.
[27] Günlük D., Lee J., Weismantel R. 2007. MINLP Strengthening for Separable Convex Quadratic
Transportation-Cost UFL. IBM Research Report: 1–16.
[28] Henrion R., Moller A. 2012. A Gradient Formula for Linear Chance Constraints under Gaussian
Distribution. Mathematics of Operations Research 37, 475–488.
[29] Henrion R., Strugarek C. 2008. Convexity of Chance Constraints with Independent Random Vari-
ables. Computational Optimization and Applications 41 (2), 263–276.
[30] Hong X., Lejeune M.A., Noyan N. 2014. Stochastic Network Design for Disaster Preparedness. IIE
Transactions. In Press.
[31] Horst R., Tuy H. 1996. Global Optimization: Deterministic Approaches. Springer, Berlin.
[32] Kataoka S. 1963. A Stochastic Programming Model. Econometrica 31 (1-2), 181–196.
[33] Kogan A., Lejeune M.A. 2013. Threshold Boolean Form for Joint Probabilistic Constraints with
Random Technology Matrix. Mathematical Programming. In Press.
[34] Küçükyavuz S. 2012. On Mixing Sets Arising in Probabilistic Programming. Mathematical Pro-
gramming 132, 31-56.
[35] Laporte G., Louveaux F.V., van Hamme L. 1994. Exact Solution to a Location Problem with
Stochastic Demands. Transportation Science 28 (2), 95–103.
[36] Larson R.C., Sadiq G. 1983. Facility Locations with the Manhattan Metric in the Presence of Bar-
riers to Travel. Operations Research 31 (4), 652–669.
[37] Lasry A., Sansom S.L., Hicks K.A., Uzunangelov V. 2011. A Model for Allocating CDC’s HIV
Prevention Resources in the United States. Health Care Management Science 14, 115–124.
[38] Latorre G., Cruz R.D., Areiza J.M., Villegas A. 2003. Classification of Publications and Models on
Transmission Expansion Planning. IEEE Transactions on Power Systems 18(2): 938–46.
[39] Lejeune M.A. 2012. Pattern-Based Modeling and Solution of Probabilistically Constrained Opti-
mization Problems. Operations Research 60 (6), 1356–1372.
[40] Lejeune M.A. 2012. Pattern Definition of the p-Efficiency Concept. Annals of Operations Research
200 (1), 23–36.
[41] Lejeune M.A., Shen S. 2013. Multi-Objective Probabilistically Constrained Programming with
Variable Risk: New Models and Applications. Working Paper. Submitted.
[42] Li X., Tomasgard A., Barton P.I. 2012. Decomposition Strategy for the Stochastic Pooling Problem.
Journal of Global Optimization 54, 765–790.
[43] Liberti L. 2006. Writing Global Optimization Software. In: Liberti L., Maculan N. (Eds.) Global
Optimization: from Theory to Implementation. Springer, Berlin, 211–262.

31
[44] LINDO. Solver Suite. Lindo Systems Inc., Chigago, IL. http://www.lindo.com/
[45] Lipsitch M., Riley S., Cauchemez S., Ghani A.C., Ferguson N.M. 2009. Managing and Reducing
Uncertainty in an Emerging Influenza Pandemic. New England Journal of Medicine 361 (2), 112–
115.
[46] Marathe M., Vullikanti A.K.S. 2013. Computational Epidemiology. Communications of the ACM
56 (7), 88–96.
[47] wpweb2.tepper.cmu.edu/fmargot/stoch.html
[48] McCormick G.P. 1976. Computability of Global Solutions to Factorable Nonconvex Programs: Part
I. Convex Underestimating Problems. Mathematical Programming 10, 147–175.
[49] Murray W., Shanbhag U.V. 2006. A Local Relaxation Approach for the Siting of Electrical Substa-
tions. Computational Optimization and Applications 33, 7–49.
[50] Perron S., Hansen P., Le Digabel S., Mladenovic N. 2010. Exact and Heuristic Solutions of the
Global Supply Chain Problem with Transfer Pricing. European Journal of Operational Research
202, 864–879.
[51] Prékopa A. 1974. Programming under Probabilistic Constraints with a Random Technology Matrix.
Mathematische Operationsforschung und Statistik 5 (2), 109–116.
[52] Prékopa A. 1990. Dual Method for a One-Stage Stochastic Programming with Random rhs Obeying
a Discrete Probability Distribution. Zeitschrift of Operations Research 34, 441–461.
[53] Prékopa A. 2003. Probabilistic Programming Models. Chapter 5 in: Stochastic Programming:
Handbook in Operations Research and Management Science 10. Eds: Ruszczyński A., Shapiro
A. Elsevier Science Ltd, 267–351.
[54] Ray J., Boggs P.T., Gay D.M., Lemaster M.N., Ehlen M.E. 2009. Risk-Based Decision making
for Staggered Bioterrorist Attacks: Resource Allocation and Risk reduction in Reload Scenarios.
Sandia Report 2009-6008, 1–61.
[55] Rawls C.G., Turnquist M.A. 2010. Pre-positioning of Emergency Supplies for Disaster Response.
Transportation Research Part B 44, 521-534.
[56] Ruszczyński A. 2002. Probabilistic Programming with Discrete Distribution and Precedence Con-
strained Knapsack Polyhedra. Mathematical Programming 93 (2), 195–215.
[57] Sahinidis N.V. 1996. Baron: A General Purpose Global Optimization Software Package. Journal
of Global Optimization 8, 201–205.
[58] Sen S. 1992. Relaxations for Probabilistically Constrained Programs with Discrete Random
Variables. Operations Research Letters 11, 81–86.
[59] Tanner M.W., Ntaimo L. 2010. IIS Branch-and-Cut for Joint Chance-Constrained Stochastic Pro-
grams and Application to Optimal Vaccine Allocation. European Journal of Operational Research
207 (1), 290–296.
[60] Tawarmalani M., Sahinidis N.V. 2002. Convexification and Global Optimization in Continuous and
Mixed-Integer Nonlinear Programming: Theory, Algorithms, Software and Aapplications. Noncon-
vex Optimization and Its Applications 65, Kluwer Academic Publishers, Dordrecht.
[61] Tawarmalani M., Sahinidis N.V. 2004. Global Optimization of Mixed-Integer Nonlinear Programs:
A Theoretical and Computational Study. Mathematical Programming 99, 563–591.
[62] van Ackooij W., Henrion R., Moller A., Zorgati R. 2011. On Joint Probabilistic Constraints with
Gaussian Coefficient Matrix. Operations Research Letters 39 (2), 99–102.
[63] Ventura J.A. 1991. Computational Development of a Lagrangian Dual Approach for Quadratic
Networks. Networks 21, 469–485.

32
Appendix
Appendix I: Illustration of Boolean Modeling Method with a Numerical Example
We use the stochastic programming problem (71) to illustrate the different steps of the Boolean reformu-
lation method:
max 3x1 + x2 + x3
2ξ12 x1 x2 + 4ξ13 x1 x3 − 24 ≤ 0 
 


 

+ 2 − 29 ≤ 0
 



 3ξ21 x 2 x 1 6ξ 22 x 2




ξ ≥ 0.7 ,
 
subject to P  x x ≤ 4 (71)
 
 12 1 2 

ξ
 



 x
13 1 3x ≤ 4 



 ξ22 x2 ≤ 3

 


2
x1 , x2 , x3 ≥ 0
where s11 = 2, s12 = s21 = 4, s13 = s23 = 0, s22 = 6. The cumulative probability distribution F of
the random variable ξ = [ξ1 , ξ2 , ξ3 ] = [ξ12 , ξ13 , ξ22 ] with ξ12 = ξ21 is given in Table 6. The realizations
ωk = [ωk1 , ωk2 , ωk3 ] of ξ are equally likely.

Table 6: Probability Distribution


k 1 2 3 4 5 6 9 8 9 10
ωk1 9 1 1 4 2 4 9 9 9 9
ωk2 10 1 3 5 3 5 8 2 8 5
ωk3 2 1 3 2 1 1 1 4 3 6
F(ωk ) 0.5 0.1 0.2 0.4 0.2 0.3 0.6 0.2 0.8 0.9

The sufficient-equivalent set of cut points is:

C e = {c11 = 9; c21 = 5; c22 = 8; c23 = 10; c31 = 3; c32 = 4; c33 = 6} .


Table 7 displays the recombinations and their binary images (i.e., relevant Boolean vectors) obtained
with C e .
Table 7: Reformulations and Relevant Boolean Vectors

Numerical Vectors Boolean Vectors Relevant Boolean Sets


k ωk1 ωk2 ωk3 βk11 βk21 βk22 βk23 βk31 βk32 βk33
11 9 5 3 1 1 0 0 1 0 0
Ω̄−B
12 9 5 4 1 1 0 0 1 1 0
9 9 8 3 1 1 1 0 1 0 0
10 9 5 6 1 1 0 0 1 1 1
13 9 8 4 1 1 1 0 1 1 0
14 9 8 6 1 1 1 0 1 1 1
15 9 10 3 1 1 1 1 1 0 0 Ω̄+B
16 9 10 4 1 1 1 1 1 1 0
17 9 10 6 1 1 1 1 1 1 1
Cut Points
9 5 8 10 3 4 6

33
Appendix II: Additional Results for Stochastic Facility Location Models

Table 8: Modification of strong branching parameters for Trb-L with formulation M2-2
(30,10) (30,5) (30,2) (20,10) (20,5) (20,2) (10,10) (10,5) (10,2)
1B2C 2 95 5000 7.8 7.5 10.6 7.8 7.9 8.7 8.6 7.2 9.3
1B2C 3 95 5000 284.1 257.7 309.3 290.4 263.1 331.3 394.5 486.0 296.8
1B2C 4 95 5000 4201.1 4145.9 4262.4 4359.8 4296.8 3868.3 4254.8 4600.0 4595.4
1B2C 2 90 5000 105.0 106.6 101.0 99.6 110.1 93.2 106.2 88.7 92.4
1B2C 3 90 5000 1707.2 1758.3 3637.5 1606.9 2233.9 2841.4 2708.0 2240.6 3001.1
1B2C 4 90 5000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 2 85 5000 281.0 350.4 694.4 349.4 392.9 587.3 403.6 940.1 705.9
1B2C 3 85 5000 7002.5 5397.4 8895.8 9000 5779.6 9000 6673.7 5794.0 9000
1B2C 4 85 5000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 2 80 5000 9000 3487.6 6666.5 4554.4 6167.1 4820.5 2883.8 5254.8 5106.3
1B2C 3 80 5000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 4 80 5000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 2 95 10000 19.0 15.7 15.7 12.5 16.3 14.0 14.0 15.8 13.7
1B2C 3 95 10000 325.7 362.1 388.9 350.7 361.1 401.2 292.3 363.7 386.8
1B2C 4 95 10000 5347.0 3205.0 3877.1 9000 3897.6 3588.3 4139.7 3683.1 4219.2
1B2C 2 90 10000 120.3 121.8 131.1 115.1 96.3 127.9 140.4 141.9 139.2
1B2C 3 90 10000 2651.0 3808.5 2629.1 2944.0 1772.6 3342.0 4151.9 1882.5 3210.1
1B2C 4 90 10000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 2 85 10000 431.8 491.0 674.7 448.3 545.9 648.0 322.9 397.3 688.6
1B2C 3 85 10000 9000 9000 9000 8362.1 9000 9000 6358.2 9000 9000
1B2C 4 85 10000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 2 80 10000 5546.1 9000 9000 2915.3 3379.4 9000 5599.6 5092.2 9000
1B2C 3 80 10000 9000 9000 9000 9000 9000 9000 9000 9000 9000
1B2C 4 80 10000 9000 9000 9000 9000 9000 9000 9000 9000 9000
Average Time 4917.9 4729.8 5095.6 4850.7 4596.7 4986.3 4602.2 4666.2 5061.0
Minimal Time 7.8 7.5 10.6 7.8 7.9 8.7 8.6 7.2 9.3
Maximal Time 9000 9000 9000 9000 9000 9000 9000 9000 9000
Unsolved Instances 10 10 11 10 9 11 8 9 11

34
Table 9: Modification of bound tightening parameters for Trb-L with formulation M2-2
yy yn ny nn
1B2C 2 95 5000 9.9 8.6 13.5 9.9
1B2C 3 95 5000 272.1 394.5 264.8 373.2
1B2C 4 95 5000 4040.7 4254.8 3982.1 3981.4
1B2C 2 90 5000 116.4 106.2 112.1 96.7
1B2C 3 90 5000 1968.4 2708.0 1950.5 2679.3
1B2C 4 90 5000 9000 9000 9000 9000
1B2C 2 85 5000 380.6 403.6 388.8 419.9
1B2C 3 85 5000 5898.3 6673.7 6112.5 9000
1B2C 4 85 5000 9000 9000 9000 9000
1B2C 2 80 5000 3713.8 2883.8 9000 9000
1B2C 3 80 5000 9000 9000 9000 9000
1B2C 4 80 5000 9000 9000 9000 9000
1B2C 2 95 10000 17.6 14.0 15.8 13.4
1B2C 3 95 10000 333.3 292.3 295.0 317.1
1B2C 4 95 10000 4164.0 4139.7 4428.4 4009.4
1B2C 2 90 10000 114.9 140.4 120.6 130.7
1B2C 3 90 10000 1893.2 4151.9 7177.4 1617.2
1B2C 4 90 10000 9000 9000 9000 9030.8
1B2C 2 85 10000 322.9 322.9 1033.2 542.6
1B2C 3 85 10000 9000 6358.2 9000 6267.4
1B2C 4 85 10000 9000 9000 9000 9000
1B2C 2 80 10000 8284.6 5599.6 9000 5480.1
1B2C 3 80 10000 9000 9000 9000 9000
1B2C 4 80 10000 9000 9000 9000 9000
Average Time 4688.8 4602.2 5203.9 4830.8
Minimal Time 9.9 8.6 13.5 9.9
Maximal Time 9000 9000 9000 9000
Unsolved Instances 9 8 11 10

35
Table 10: Modification of strong branching parameters for Trb-I on formulation M2-2-MC1-1.
(30,10) (30,5) (30,2) (20,10) (20,5) (20,2) (10,10) (10,5) (10,2)
2B1C 2 95 5000 7.5 4.5 4.5 4.3 3.8 3.8 4.7 5.3 5.0
2B1C 3 95 5000 37.9 46.3 41.5 45.0 23.8 52.5 76.0 49.6 29.8
2B1C 4 95 5000 154.1 289.0 206.9 501.5 235.5 131.9 363.3 108.8 418.6
2B1C 2 90 5000 15.5 8.0 9.6 8.8 11.3 8.9 8.6 9.1 10.1
2B1C 3 90 5000 73.7 59.6 96.7 109.0 195.3 41.5 146.6 84.7 77.3
2B1C 4 90 5000 365.2 419.6 581.2 733.9 332.4 160.1 704.2 298.0 384.4
2B1C 2 85 5000 43.0 58.2 41.6 57.6 34.4 27.2 24.5 43.0 28.5
2B1C 3 85 5000 3545.4 321.5 227.4 415.5 745.7 425.1 427.6 287.3 388.3
2B1C 4 85 5000 911.3 1807.2 4391.9 1418.1 506.8 2123.6 6154.2 1037.7 835.5
2B1C 2 80 5000 140.6 205.2 146.2 124.8 163.8 129.1 152.1 140.6 56.7
2B1C 3 80 5000 1160.9 3222.3 2032.1 1065.9 1460.5 9000 4404.0 7009.6 1239.8
2B1C 4 80 5000 1402.5 3949.7 2887.7 6854.9 8564.5 4454.0 5934.5 3420.4 1617.0
2B1C 2 95 10000 8.8 7.5 5.2 4.2 3.8 5.2 4.9 5.0 5.1
2B1C 3 95 10000 58.9 54.9 32.4 47.1 62.3 36.9 52.5 43.2 48.1
2B1C 4 95 10000 159.2 407.1 223.9 358.6 557.3 230.9 432.0 220.3 160.5
2B1C 2 90 10000 17.2 11.9 9.5 11.4 10.4 9.9 16.2 11.7 10.0
2B1C 3 90 10000 74.2 148.5 104.3 126.9 1213.7 114.8 115.1 1832.9 60.9
2B1C 4 90 10000 395.5 672.2 177.3 1164.8 1128.4 1149.7 958.2 253.8 371.0
2B1C 2 85 10000 41.5 79.2 32.9 73.3 38.2 23.7 50.7 43.6 29.5
2B1C 3 85 10000 455.8 667.2 837.2 908.6 329.2 9000 328.9 252.6 329.3
2B1C 4 85 10000 575.0 2198.5 3172.2 4160.1 2765.4 1361.6 4571.8 1453.2 567.3
2B1C 2 80 10000 174.4 238.2 176.5 209.6 145.8 130.3 126.2 95.2 110.9
2B1C 3 80 10000 1328.0 1136.4 2106.7 1760.0 831.5 9000 1540.6 2399.9 1507.3
2B1C 4 80 10000 1600.5 2090.9 9000 6995.7 5932.9 1078.5 7864.7 9000 1609.6
Average Time 531.1 754.3 1106.1 1131.6 1054.0 1612.5 1435.9 1171.0 412.5
Minimal Time 7.5 4.5 4.5 4.2 3.8 3.8 4.7 5.0 5.0
Maximal Time 3545.4 3949.7 9000 6995.7 8564.5 9000 7864.8 9000 1617.0
Unsolved Instances 0 0 1 0 0 3 0 1 0

36
Table 11: Modification of strong branching parameters for Trb-L-I on formulation M2-2-MC1-1.
(30,10) (30,5) (30,2) (20,10) (20,5) (20,2) (10,10) (10,5) (10,2)
2B1C 2 95 5000 5.5 5.4 4.7 4.8 3.1 4.4 6.2 5.8 4.2
2B1C 3 95 5000 71.2 35.2 40.7 61.0 24.0 25.8 73.9 48.2 34.1
2B1C 4 95 5000 352.1 325.6 146.0 505.8 391.7 178.3 355.3 114.7 354.4
2B1C 2 90 5000 9.5 9.5 10.6 9.4 13.5 8.8 9.0 9.0 11.6
2B1C 3 90 5000 132.3 58.3 68.4 111.9 128.1 51.5 230.5 77.9 68.4
2B1C 4 90 5000 979.6 590.2 379.8 730.3 328.4 163.4 804.6 4580.8 795.9
2B1C 2 85 5000 71.4 54.0 47.9 53.9 32.5 20.5 24.4 37.7 25.5
2B1C 3 85 5000 641.5 282.4 346.7 369.2 1572.4 495.0 430.4 558.4 359.8
2B1C 4 85 5000 2701.6 3836.3 590.8 3568.1 563.0 1203.3 7186.6 1097.7 1023.2
2B1C 2 80 5000 260.6 158.6 769.0 140.1 177.1 96.5 129.4 115.5 61.9
2B1C 3 80 5000 3687.6 3174.0 2225.5 1955.0 1315.1 9000 1507.1 1737.9 1578.0
2B1C 4 80 5000 9000 9000 1513.0 7900.7 7760.7 2213.8 7100.8 2593.2 1982.3
2B1C 2 95 10000 9.5 7.8 6.1 4.3 3.9 5.2 4.1 5.3 4.7
2B1C 3 95 10000 113.5 51.5 30.6 47.7 57.5 35.9 53.9 40.0 49.9
2B1C 4 95 10000 578.9 424.6 199.3 379.9 531.0 231.0 353.0 272.4 140.4
2B1C 2 90 10000 12.8 11.1 10.9 11.3 10.7 10.3 13.0 15.6 11.8
2B1C 3 90 10000 99.0 135.9 174.4 134.0 1906.3 67.8 126.0 68.5 62.9
2B1C 4 90 10000 965.2 580.0 696.8 905.1 472.6 190.8 784.9 257.1 368.0
2B1C 2 85 10000 85.7 145.2 45.2 46.6 33.9 27.5 67.2 42.2 29.3
2B1C 3 85 10000 986.1 525.9 469.5 794.0 735.2 3177.4 386.6 289.2 307.6
2B1C 4 85 10000 7269.7 1152.0 9000 5912.0 2448.6 1296.4 5166.7 2763.0 581.6
2B1C 2 80 10000 322.9 253.4 788.0 134.6 133.0 142.3 122.1 93.5 134.7
2B1C 3 80 10000 1723.7 936.2 4075.0 3199.8 1061.7 940.8 1588.8 2480.2 6233.4
2B1C 4 80 10000 7717.3 4100.0 2320.8 8721.3 9000 1350.9 9000 7389.9 1542.7
Average Time 1574.9 1077.2 998.3 1487.5 1196.0 872.4 1480.2 1028.9 656.9
Minimal Time 5.5 5.4 4.7 4.3 3.1 4.4 4.1 5.3 4.2
Maximal Time 9000 9000 9000 8721.3 9000 9000 9000 7389.9 6233.4
Unsolved Instances 1 1 1 0 1 1 1 0 0

37
Variations of the bound tightening options of Couenne, namely
• feasibility bt: yes or no
• optimality bt: yes or no.
This gives us 4 parameter settings, labeled “yy”, “yn”, “ny”, and “nn”, the first letter referring to the
setting of feasibility bt.

Table 12: Modification of bound tightening parameters for Trb-I with formulation M2-2-MC1-1.
yy yn ny nn
2B1C 2 95 5000 9.9 5.0 8.1 4.0
2B1C 3 95 5000 65.2 29.8 59.8 28.5
2B1C 4 95 5000 160.5 418.6 159.6 414.5
2B1C 2 90 5000 15.0 10.1 15.7 9.5
2B1C 3 90 5000 87.9 77.3 73.6 73.7
2B1C 4 90 5000 393.5 384.4 381.9 373.3
2B1C 2 85 5000 42.1 28.5 43.6 26.1
2B1C 3 85 5000 447.5 388.3 356.1 391.9
2B1C 4 85 5000 1137.6 835.5 620.9 1098.2
2B1C 2 80 5000 138.4 56.7 153.0 56.0
2B1C 3 80 5000 1192.5 1239.8 1531.9 924.0
2B1C 4 80 5000 1328.6 1617.0 1677.1 1283.6
2B1C 2 95 10000 8.8 5.1 1727.9 4.4
2B1C 3 95 10000 60.1 48.1 8.2 47.3
2B1C 4 95 10000 173.0 160.5 64.9 153.3
2B1C 2 90 10000 15.7 10.0 167.6 9.9
2B1C 3 90 10000 75.2 60.9 15.2 56.8
2B1C 4 90 10000 379.1 371.0 84.8 362.6
2B1C 2 85 10000 42.6 29.5 384.3 23.1
2B1C 3 85 10000 390.6 329.3 38.0 313.7
2B1C 4 85 10000 580.1 567.3 397.7 541.9
2B1C 2 80 10000 156.0 110.9 808.0 121.2
2B1C 3 80 10000 952.6 1507.3 106.1 1178.5
2B1C 4 80 10000 1664.4 1609.6 1319.2 1536.1
Average Time 418.2 424.3 423.4 390.3
Minimal Time 9.9 5.0 8.1 4.0
Maximal Time 1328.6 1617.0 1677.1 1283.6
Unsolved Instances 0 0 0 0

38
Table 13: Modification of bound tightening parameters for Trb-L-I with formulation M2-2-MC1-1.
yy yn ny nn
2B1C 2 95 5000 7.5 4.2 8.3 28.6
2B1C 3 95 5000 37.9 34.1 33.5 327.8
2B1C 4 95 5000 154.1 354.4 154.4 9.0
2B1C 2 90 5000 15.5 11.6 16.6 62.3
2B1C 3 90 5000 73.7 68.4 86.3 752.8
2B1C 4 90 5000 365.2 795.9 371.4 26.0
2B1C 2 85 5000 43.0 25.5 40.6 3328.0
2B1C 3 85 5000 3545.4 359.8 350.3 936.5
2B1C 4 85 5000 911.3 1023.2 804.7 59.6
2B1C 2 80 5000 140.6 61.9 86.6 890.0
2B1C 3 80 5000 1160.9 1578.0 1184.1 1311.5
2B1C 4 80 5000 1402.5 1982.3 9000 4.0
2B1C 2 95 10000 8.8 4.7 9.2 47.1
2B1C 3 95 10000 58.9 49.9 57.9 133.5
2B1C 4 95 10000 159.2 140.4 152.5 10.6
2B1C 2 90 10000 17.2 11.8 15.1 59.0
2B1C 3 90 10000 74.2 62.9 77.2 349.2
2B1C 4 90 10000 395.5 368.0 391.0 23.0
2B1C 2 85 10000 41.5 29.3 47.4 246.2
2B1C 3 85 10000 455.8 307.6 483.8 540.1
2B1C 4 85 10000 575.0 581.6 605.1 119.7
2B1C 2 80 10000 174.4 134.7 154.9 1353.8
2B1C 3 80 10000 1328.0 6233.4 5830.7 1577.0
2B1C 4 80 10000 1600.5 1542.7 1676.1 1577.0
Average Time 654.8 524.9 1011.4 644.7
Minimal Time 7.5 4.2 8.3 4.0
Maximal Time 3545.4 1982.3 9000 3328.0
Unsolved Instances 0 0 1 0

39

Potrebbero piacerti anche