Results

Tour

Fitness

Download

Genetic Algorithms are used where deterministic ones fail - or just for fun. This practical approach shows the example of the travelling salesman. Given a set of say 20 cities and the distances between, what's the fastest tour to reach all destinations? Not simple with a procedural approach... However, a genetic algorithm solves this in a few seconds. 
The program presented here is written in C (by R. Cattin) and can be adapted to other problems. A example set of 20 cities was chosen with random coordinates in the range of 0..1. The initialization of the program generates 100 random tours through this 20 places. This tours are then improved over 200 generations. The improvement is acheived by crossover, selection and mutation. Selection is done by a random roulette-wheel function according to the fitness of the tours. The fitness of a tour corresponds to the distance to travel, of course. 
And: It's not all written here, explore it!
 

Results	Result tables of a Travelling Salesman example.
Tour	Shows tour through the 20 cities (random tour and improved by genetic algorithm).
Fitness	Shows how the distance the salesman has to travel develops over time (corresponds to the fitness of the population).
Download	Download all you like of this demo.

Extension of the course "Neuronal Networks" thought by W. Businger at the University of Applied Sciences Biel (HTA Biel). Notice that the example program in C is written and copyrighted by R. Cattin, Professor at HTA Biel.