This project uses neural network and genetic algorithm to optimize the cat's behavior. In each generation, the last cat flying gets to reproduce thus keeping its gene in the simulation. Every reproduced cat will be given the parent's gene that has slightly mutated and once in a while it happens to be superior than the predecessor.
This is a test to see whether scratch cat is really learning under any type of obstacle. With this one, It somehow prefers collision over victory. After 30 generations it's highest score is 2. Better to run it in turbowarp.
