Even when looking at something as barebones as perceptrons, it's hard to really see "why" they work. Check out the Karpathy "Zero to Hero" videos, and try to f

Even when looking at something as barebones as perceptrons, it's hard to really see "why" they work. Check out the Karpathy "Zero to Hero" videos, and try to follow along by building an MLP implementation in your own language of choice. He does a good job of building intuition because he doesn't skip much of anything.