Like, logistic regression uses a non-linearity to convert the outputs to the 0-1 scale. How does that differ from a one layer neural network?

I think this would probably be a more profitable discussion if you could define Machine Learning for me.