Ever since I learned about Darwin’s theory of evolution by natural selection, I have wanted to create an artificial environment in which the forces of evolution could work as freely as possible. Specifically, I wanted to see if I could make an image generator that could produce constantly new, beautiful and interesting images that didn’t seem confined to a certain look.
Having stumbled across the idea of genetic programming (GP), I wrote a Java application that evolves programs that construct images. In GP, programs compete in tournaments, where winners get to reproduce. During reproduction, programs are altered by crossover and mutation. In practice, the application shows me a few images that are created by programs which are randomly drawn from a population. When I click one of the images, the application lets the program that created that image reproduce in the population. The process is repeated. In theory, since I each time click the image I find to be the most interesting, the interestingness of the images will increase over time.
Having stumbled across the idea of genetic programming (GP), I wrote a Java application that evolves programs that construct images. In GP, programs compete in tournaments, where winners get to reproduce. During reproduction, programs are altered by crossover and mutation. In practice, the application shows me a few images that are created by programs which are randomly drawn from a population. When I click one of the images, the application lets the program that created that image reproduce in the population. The process is repeated. In theory, since I each time click the image I find to be the most interesting, the interestingness of the images will increase over time.
No comments:
Post a Comment