|
|
|
Lecturer: Cor Hurkens Optimization is finding the best of all possible solutions within a certain model of a situation. For some models an optimal solution can be found in reasonable time with the use of standard techniques. Quite often though, models do not permit an “easy” solution so the mathematical engineer must find a balance between computing time and solution quality. He must know some generally applicable techniques, use existing software to implement them and come up with creative combinations of techniques. The models of industrial mathematics can be continuous and linear, combinatorial, or highly nonlinear. Models close to realistic situations may be very complex and mathematically intractable. Optimization means either getting the true optimal solution, i.e., the cheapest among all alternatives, or a solution that comes close to the true optimum. In the first case, one may be dealing with a problem for which an existing technique will give an optimal solution in little computing time or one may be tackling an intrinsic difficult problem for which finding the optimum is very time-consuming and clever engineering is needed. In the second case, one may settle for a sub-optimal solution. There are fast (construction) heuristics that build a solution greedily with varying solution quality. There are also slow heuristics, mainly based on local search, attempting to improve upon a certain solution The course is divided into two parts. In the first part all trainees participate, in the second part only the trainees of the B- and C/I-orientation. The topics covered in the first part are:
|