Heuristic Search

According to the basic strategies of heuristic search algorithms not the absolute optimal solution is found but - more likely - one of the less optimal ones because there is no possibility to assure the existance of an distinct optimal solution. The randomly selecting starting point is the reason that repeated searches yield different results.

To evaluate the performance of a seach algorithm the calculation time and technoligical eligibility has to be taken into account: a good algorithm finds an optimal solution in reasonable time.

To evaluate the performance of the search algorithms each possible solution in an reasonable large search room was calcuated; the tabu and the genetic search each where allowed to search 300 times in this search room.

The frequency distributions of all sollutions in the search room where compared to the solutions found by the tabu and the genetic search (see figure). The genetic search algortithm as well as the tabu search show good performance and convergence. Convergence is in this case defined as the fact that the search algorithms find solutions that are elements of the "best" solutions in the frequency distribution of the complete search room.

Both search algorithms drastically reduce computing time (1 hour vs. 72 hours) at good performance.

To top