Paralel Genetic Algorithm Visualization

this genetic algorithm searches for the solution of the game marble solitaire (http://www.c-sharpcorner.com/UploadFile/deskenny/MarbleSolitaire11292005013450A­M/Images/marble.gif) where the point is to remove as much marbles as possible by jumping over them.

The algorithm uses a multi-island system, where the islands are merged to a 4*5 grid, and the neighbors connected. In every 10 cycle the best individual copies himself over the worst one on the neighboring island. Every island has 20 individual population. This is okay for demonstration purposes because the refreshing of the data will be not frequent enough to see how the algorithm works.

There is also a chance of reinitialization, in chance 1:3, when the whole population of a specific island gets reinitialized (the island selection is also random).

the fitness function is the reaming marbles on the table. every gene set of an individual contains a specific step, what is sequential applied from the individual to the playground (the marble table).
Steps are organized with three attributes: X,Y coordinate of the marble and a direction of the jump. The solution has 31 steps (because this is the amount of marbles to remove) so a gene of an individual contains 93 data long vector.
The problem is that more close we are to the global best (solution) more strong are the local bests (the pseudo solutions).
if you have any question, feel free to contact me at: erik at cstudios dot sk.

Category:

Science & Technology

Tags:

• Paralel
• Genetic
• Algorithm
• Fitnes
• Visualization
• STU
• FEI
• URPI
• Erik
• Putz
• Matlab