23.06.2013 Views

Implementation of Cutting Plane Separators for Mixed Integer ... - ZIB

Implementation of Cutting Plane Separators for Mixed Integer ... - ZIB

Implementation of Cutting Plane Separators for Mixed Integer ... - ZIB

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

44 Chapter 4. <strong>Cutting</strong> <strong>Plane</strong> Separator <strong>for</strong> the 0-1 Knapsack Problem<br />

33, 12, 51, 58]. This procedure involves the solution <strong>of</strong> a knapsack problem in every<br />

lifting step. Zemel [61] has introduced a polynomial algorithm which uses dynamic<br />

programming to solve these knapsack problems. Another type <strong>of</strong> lifting is called<br />

down-lifting, where variables fixed at their upper bounds are lifted. The idea <strong>of</strong><br />

sequential down-lifting has been introduced in [59], where it was used to strengthen<br />

canonical inequalities.<br />

The inequalities derived by sequential lifting, in general, depend on the sequence<br />

in which the variables are lifted. A simultaneous up-lifting procedure to strengthen<br />

minimal cover inequalities was studied by Balas and Zemel [12], but its computational<br />

burden prevents it from being applied in practice. In [60], the property <strong>of</strong><br />

superadditivity <strong>of</strong> the lifting function has been explored which leads to sequence<br />

independent lifting. Building on the results <strong>of</strong> [60], Gu, Nemhauser and Savelsbergh<br />

[32] and Atamturk [5] have investigated the concept <strong>of</strong> superadditive up-lifting to<br />

strengthen minimal cover inequalities. Here, the lifting function which, in general, is<br />

not superadditive is approximated by a so-called superadditive valid lifting function<br />

to obtain sequence independent lifting (see Section 2.2).<br />

The results <strong>of</strong> the theoretical study <strong>of</strong> the 0-1 knapsack polytope have been used<br />

in linear programming based branch-and-cut algorithms to solve IPs and MIPs.<br />

Crowder, Johnson and Padberg [22] pioneered the use <strong>of</strong> lifted inequalities and<br />

successfully solved several instances <strong>of</strong> IPs which were, at the time, considered to<br />

be unsolvable. They separated the class <strong>of</strong> lifted minimal cover inequalities using<br />

sequential up-lifting and the class <strong>of</strong> lifted (1,k)-configuration inequalities using sequential<br />

up-lifting. Since then, there have been several other successful applications<br />

<strong>of</strong> lifted valid inequalities <strong>for</strong> the 0-1 knapsack polytope. Van Roy and Wolsey [55]<br />

separated the class <strong>of</strong> lifted cover inequalities using sequential up- and down-lifting.<br />

H<strong>of</strong>fman and Padberg [35] and Gu, Nemhauser and Savelsbergh [30] implemented<br />

a cutting plane separator which separates the class <strong>of</strong> lifted minimal cover inequalities<br />

using sequential up- and down-lifting (LMCI1). In [30], a computational study<br />

was presented in which many <strong>of</strong> the algorithmic and implementation choices were<br />

evaluated which have to be made when implementing this cutting plane separator.<br />

Especially, it turned out that using both up-lifting and down-lifting instead <strong>of</strong> using<br />

only up-lifting leads to a better per<strong>for</strong>mance <strong>of</strong> the cutting plane separator. Martin<br />

[44] separated the class <strong>of</strong> lifted extended weight inequalities using sequential upand<br />

down-lifting (LEWI).<br />

For our cutting plane separator, we have to decide which <strong>of</strong> the above mentioned<br />

classes <strong>of</strong> valid inequalities we want to separate. To our knowledge, no paper has<br />

been published presenting computational results <strong>for</strong> separating the class <strong>of</strong> lifted<br />

minimal cover inequalities using superadditive up-lifting (LMCI2). Besides, we know<br />

<strong>of</strong> no paper in which the per<strong>for</strong>mance <strong>of</strong> a separation algorithm <strong>for</strong> the class <strong>of</strong><br />

LMCI1 has been compared to that <strong>of</strong> a separation algorithm <strong>for</strong> the class <strong>of</strong> LEWI.<br />

Thus, it is not clear, which <strong>of</strong> these three classes <strong>of</strong> valid inequalities leads to a best<br />

per<strong>for</strong>ming cutting plane separator. There<strong>for</strong>e, we investigate separation algorithms<br />

<strong>for</strong> all three classes <strong>of</strong> valid inequalities in this chapter.<br />

In Section 4.2, we give a brief introduction to these classes <strong>of</strong> valid inequalities,<br />

and in Section 4.3, we discuss different algorithmic aspects <strong>of</strong> the corresponding separation<br />

algorithms. In Section 4.4, we present a computational study. It evaluates<br />

the effect <strong>of</strong> using the different algorithmic and implementation choices described

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!