This program attempts to simulate biological cell growth. Since the rules that dictate the growth are completely random, some growths look pretty random and weird. I think the next step is to subject the rules to some kind of objective function and optimize a rule set for it. But it's not very easy to qualitatively describe what constitutes a "good" grown shape. Symmetric? Convex? ...
But anyway, in current version:
- Cells have continuous internal variables that encode their type.
- Neighboring cells can sense these variables (modeling communication)
- Cells have telomeres, and can undergo apoptosis
- There is a DNA with a genome that encodes local rules on how cells behave when surrounded by other cells.
- all growth stems from a single "embryo" cell
The code can be browsed here: https://gist.github.com/1561744
A compiled .exe version can be downloaded here: http://cs.stanford.edu/~karpathy/tissuegrowth.zip
Very impressive. I want to go to stanford...
THEZAKATAK123 1 month ago