1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
|
\begin{enumerate}
\item Q: Cross products. In some of the examples you define the cross
product of two sets explicitly, while in other cases you don't (see
for instance the sets $C$ and $CxC$ in the two formulations of the
\emph{$n$ queens problen}, in
Subsection~\ref{ssec:example:n-queens-problem}). Is there any
difference in performance either way?\\
A: Basically not, as long as there is no \code{with} statement or
other expression that adds or removed elements.
If in doubt, the explicit formulation should never be slower.
%\item Q: Are indices scoped? If so, how?\\
% A:
%\item Q Are recursive function allowed, or, what is the output of
%\begin{verbatim}
% set A := { 1 .. 2};
% defnumb f(i) := i + sum <j> in A | j > i: f(j);
% defnumb g(i) := sum <j> in A | j > i: f(j) + i;
% do print f(1);
% do print g(1);
%\end{verbatim} %thanks, Uli Dorndorf
% and why isn't it what one would expect?\\
% A:\textcolor{red}{ *** THORSTEN *** }
\item Q: Is the use of extended constraints (\code{vif}, \code{vabs},
Subsection~\ref{ssec:ug:constraints}) a good idea? \\
A: The short answer is \emph{no}. The longer answer is still
\emph{no}, but perhaps more comprehensively so: Extended constraints
are linearized into the notorious \emph{big-$M$}\footnote{$M$ is a
shorthand for \emph{mistake}} constraints together with some
additional binary variables which usually deteriorate the value of
your linear programming relaxation, and that's bad. To add insult to
injury, in the case that the \emph{boolean constraint} in the
definition of an extended constraint happens to be a binary variable
\zimpl happily generates another binary variable instead of
recycling the perfectly suitable existing one, with unfortunate
consequences for the efficiency of the model. Later versions of
\zimpl might fix this.
\end{enumerate}
|
