This is the final project I did for a Neural Networks class.
We were given a simulated 2d graphical world and an avatar with a number of sensory inputs. These input values can be read at each tick of the simulation. These senses included a value that represented life, visual input, and auditory input. The project was to develop a series of neurons that read in this sensory input and control the behavior of the avatar. The life value continually decreased as a function of time and velocity. When the life value reached zero, the avatar died. The goal was train a neural network that acted as a brain for the avatar. The only thing passed along between simulation runs was the brain of the avatar, so over time, it learned to survive better.