Dendrites and neurons → dendrites are the input, the axon calculates the weight of the response (similar to how the brain interprets the inputs) → synapse.
There is a base threshold that the impulse has to pass in order for the message to be passed on.
We set these biases of which information is more important by added a set value to it.
The weights of the range of inputted data ends up looking something similar to a sigmoidal curve (which is a curve that looks like a 's').
There are hundreds of levels in the network — from an input level to the output level. Each level adds some sort of bias and change in the weight of the data.
To train a neural network you need to feed it with data, but at the right amount. Overfitting, when you train a network to the point that it can't interpret other data, is also an issue, the model becomes too specialized for a specific thing.
The NEAT algorithm is a complex genetic evolutionary form of AI. We look at the math of successful "birds," and create more with similar "genetics". This is better than typical genetic models because it also adds mutations and it does not immediately dispose mutations in case they evolve into something better.
The Flappy Bird model includes basic inputs and outputs. The model does not learn anything or "evolve" and improve. We need to find a way to judge which birds are the best or how to utlize cycles, which will enable us to understand how to make things better.