The Transportation Problem

In Chapter 7 we compared different GA approaches for handling constraints. It seems that for a particular class of problems (like the transportation problem) we can do better: we can use a more appropriate (natural) data structure (for a transportation problem, a matrix) and specialized genetic operators which operate on matrices. Such an evolution program would be much stronger method than GENOCOP: the GENOCOP optimizes any function with linear constraints, whereas the new evolution program optimizes only transportation problems (these problems have precisely n + k −1 equalities, where n and k denote the number of sources and destinations, respectively; see the description of the transportation problem below). However, it would be very interesting to see what can we gain by introducing extra problem-specific knowledge into an evolution program.