lorenz lachauer at msa

today we had the pleasure of listening to a lecture from lorenz lachauer, phd student and research assistant at the chairs of prof. dr. philippe block and prof. dr. joseph schwartz from the eth zürich. he gave an introduction to the world of graphic statics, showed his latest projects and completed the day with a grasshopper/galapagos-tutorial for the students. a big thank you!

genetic algorithm II

the second run, now with double number of random points (30). each generation produces now 25 copies, 4 points are exchanged randomly and the program stops when there is no progress over 500 generations.

the statistics:
starting length: 1744.33
optimized length: 600.37
reducing length to: 34.42 %
best solution after 1404 generations
44 steps of transformation total

genetic algorithm I

the first time I wrote a genetic algorithm. starting with a simple geometrical problem (travelling salesman - some random points, find the shortest route trough all of them) the program does the following:

- generate a random line through all of the points, measure the length
- make 10 copies and exchange 2 points with each other randomly, measure the length
- the shortest (the fittest) is the "father" for the new generation
- repeat until there are no fitter copies for 50 generations

these are the statistics for the following example:
starting length: 735.08
optimized length: 344.35
reducing length to: 46.85 %
best solution after 38 generations
16 steps of transformation total

it is fascinating to give a certain intelligence to your scripts. though we have a very simple problem the process of finding a solution to it is very promising. it is a simulation of evolution, only the fittest members can reproduce themselves...