A network engineer’s notebook, from a toy example to a router crash predictor
I have spent years reading interface counters, chasing flaps, and writing scripts that check things. So when I decided to actually understand neural networks, not just import them, I did not start with theory. I started with three tiny scripts I could run line by line and check every number by hand. This post is those three scripts, in order, with the real outputs and a picture for each one.
By the end you will have seen a neural network predict, learn, and then judge things it has never seen, all on examples from our own world: routers that are about to fall over.