Trained Automata

Alexander Gross
Custom Application, Processing
November, 2007

Back to Projects
Trained Automata

In code I constructed a program that would read a grayscale images and create a cellular automata ruleset based on evaluating their consecutive horizontal scan lines, to see which color pixel was most common throughout the images based of all the possible rule bases, for a 16 value 2D cellular automata. After the rule set was constructed. Each horizontal scan line was used as a base line to generate a new image using the newly created CA ruleset to attempt to probabilistically reconstruct some portion of the original image, even if only for a very some number of generations. Some of the scan line bases generated very simple orderly patterns, several begin to display some of the traditional characteristics of complex 2D cellular automata, as well their were a few that did a decent job of continuing the image. The image used as a base was one Abraham Lincoln.