Operations Research Letters 45 (2017) 513–518
Contents lists available at ScienceDirect
Operations Research Letters
journal homepage: www.elsevier.com/locate/orl
Lifting of probabilistic cover inequalities
Seulgi Joung, Sungsoo Park *
Department of Industrial and Systems Engineering, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea
article info
Article history:
Received 19 July 2016
Received in revised form 10 March 2017
Accepted 2 August 2017
Available online 12 August 2017
Keywords:
Chance-constrained optimization
Knapsack problem
Probabilistic cover
Lifting
abstract
We consider a chance-constrained binary knapsack problem where weights of items are independent and
normally distributed. Probabilistic cover inequalities can be defined for the problem. The lifting problem
for probabilistic cover inequalities is NP-hard. We propose a polynomial time approximate lifting method
for probabilistic cover inequalities based on the robust optimization approach. We present computational
experiments on multidimensional chance-constrained knapsack problems. The results show that our
lifting method reduces the computation time substantially.
© 2017 Elsevier B.V. All rights reserved.
1. Introduction
We consider a binary knapsack problem with uncertain weights
of items. The binary knapsack problem is to select items to
maximize the sum of profits while satisfying knapsack capacity
constraint. The binary knapsack problem without uncertainty is
known to be NP-hard [13]. Many practical problems can be for-
mulated using the binary knapsack constraint, and studies on the
knapsack solution set can be applied to any mixed integer pro-
gramming problems containing the knapsack constraint. Therefore
the binary knapsack problem has been extensively studied [13,17],
and much research is focused on the polyhedral aspects of the
binary knapsack solution set.
In applications of optimization, an important issue is how to
deal with data uncertainty. Representative approaches to deal with
data uncertainty are stochastic programming and robust optimiza-
tion. Stochastic optimization considers a probabilistic distribution
of uncertain data [6], and chance-constrained programming [8]
is one of the useful approaches of stochastic programming. It
finds the best solution satisfying constraints with probability at
least a given threshold. The chance-constrained knapsack problem
has been studied by many authors. Goel et al. [9] proposed a
polynomial time approximation scheme (PTAS) where weight of
each item has a Poisson or exponential distribution. Kleinberg
et al. [14] assumed that each weight has a Bernoulli distribution
and presented an approximation algorithm. Klopfenstein et al. [15]
proposed an approximation algorithm using robust optimization
techniques. Goyal et al. [10] presented a PTAS using a parametric
LP reformulation when item sizes are independent and normally
*
Corresponding author.
E-mail addresses: sgjoung@kaist.ac.kr (S. Joung), sspark@kaist.ac.kr (S. Park).
distributed. Han et al. [12] proposed an efficient pseudopolynomial
algorithm based on the robust optimization approach for finding
a good upper bound on the optimal value when weight of each
item has a normal distribution independent of the other items. This
paper also focuses on the chance-constrained knapsack problem
where weights are independent and normally distributed.
Cover inequalities are well-known valid inequalities for the
binary knapsack solution set. Commercial optimization softwares
such as CPLEX and Xpress-MP use cover inequalities in their
branch-and-cut algorithms. Lifting is a powerful technique to
strengthen cover inequalities [4,21]. Cover inequalities also can be
defined for the knapsack problems considering data uncertainty.
Klopfenstein et al. [16] defined robust cover inequalities for the
robust knapsack problem that is formulated using the uncertainty
set of Bertsimas et al. [5]. Song et al. [20] considered the chance-
constrained binary packing problem and proposed a problem for-
mulation using probabilistic covers. They focused on the case
with finite number of scenarios and considered an approximate
lifting method. The chance-constrained knapsack problem is a
special case of the chance-constrained packing problem. Atamtürk
et al. [3] described cover inequalities for the submodular knapsack
set and proposed a polynomial time approximate lifting algorithm
using parametric LP. Note that the chance-constrained knapsack
problem with independent and normally distributed weights is
one of submodular knapsack problems. Atamtürk et al. [1] in-
vestigated separation and extension of cover inequalities for the
conic quadratic knapsack constraint with generalized upper bound
(GUB) constraints.
In this paper we propose a polynomial time approximate lifting
algorithm for cover inequalities for the chance-constrained knap-
sack problem. We assume that weights of items are independent
and normally distributed. We formulate the subproblem for lifting
http://dx.doi.org/10.1016/j.orl.2017.08.006
0167-6377/© 2017 Elsevier B.V. All rights reserved.