>>11265062I'll give you some advice.
I've tried something similar, but i have a neuroscience background.
You have two types of arrays. One kind represents neurons current state, on or off.
The other array represents the magnitude of the effect one neuron has on another.
So for example.
Neuron1 x Effect1 + N2E2 + N3E3 + ... NnEn = TotalTargetAlpha
If TotalTargetAlpha >= 1, the NeuronAlpha is turned on, and it's value gets set to 1. Effect values range from 1(pro-action potential in neuron alpha) to -1 (anti-action potential in neuron alpha), where 0 corresponds to a lack of synapse between NeuronK and neuron alpha. Effectively, you could have a one dimensional array full of neurons whose state is 8 bit, or however many bits for however many decimal places you want to approximate.
Then an array to signify connections or effects. If you have your neuron array with n*1 components, your connection array has n*n components, where [x,y] in the connection/effect array works thusly.
Neuron(y) = sum(Neuron(x)*E(x)) for all x in neuron array.
You could have an input and output layer too, and allow some technique for the network to determine its own effect array values, which are optimized to perform some task.
It'd essentially be a single hidden layer neural network, where every connection is feedneutral ie it feeds into the same layer, or feeds output into the external, or takes inputs from the external
