Decropixel is an artificial intelligence which draws simple pixel art and uses feedback to improve its neural network.
The neural network contains an input layer, an output layer and some layers in between. Layers are no more than arrays of neuron structures. Each neuron has a connection to every other neuron in the next and previous layers, if these exist.
Neurons can be stimulated with a certain level of stimulus. When a neuron is fired, it passes its stimulation to all the neurons it has a connection to in the next layer. However, passing the stimulation value alone does not provide the artificial intelligence with the ability to learn. In fact, if the stimuli were the only fuel powering the behaviour of the neural network, for the same stimuli, it would always produce the same behaviour. Therefore, this is where the other factors come into play.
First, there are connection strengths. These will determine how much of the stimulus will actually be passed onto the next neuron. Each connection has a certain strength associated with it, which can be adjusted to be stronger or weaker, depending on the feedback given to the artificial intelligence.
Then, there are biases. Each neuron has a certain bias, which represents a value that is added or subtracted from the stimulus (depending on whether it's positive or negative). These can also be adjusted with relation to the feedback received.
The neurons, their biases and the connections they form between each other form the neural network itself. A simple, but complex, system that allows a machine to learn.
The license must be included with
each distribution of this software
in the form of a LICENSE file.