Neurons can actually simulate logic gates.
Neuron basically behaves like this (simplified):
`output = exp{ sum(weight_k * input_k) }, if (output>max_output) then (output = 0 for T ms)`
You can make NOT as `exp(-1*input_1)`. You can make AND by picking weights in such a way, that `exp(w)` is a small number, but exp`(2*w)` is big. (And you can use second neuron to make them even more distinct). With AND and NOT you can make NAND, and from NAND – everything else.
Latest Answers