Genetic Algorithm Paints Mona Lisa
Slashdot has a link to an interesting creative computation project. A genetic algorithm is used to create an approximation of the Mona Lisa. It starts with a random array of variables, uses them to draw a bunch of polygons, and tests the result against a picture of the Mona Lisa. If it thinks that the random polygon image looks a bit like the Mona Lisa (using a similarity measure – hint – this is not as easy as it sounds, but not too difficult either) it modifies it a bit and tests it again against the picture. If the new modified version looks more like the Mona Lisa than the previous version, it stores the new variables, modifies them, and tests again.
After just under 1 million iterations, you’d be surprised how good it gets.