Artificial Intelligence for Humans, Vol 3: Deep Learning and Neural Networks

http://karpathy.github.io/neuralnets/

The best thing about this article is how he breaks down the math. The main idea: slightly tweak an argument to a function to get the output slightly closer to your desired output. He then presents a couple of ways to achieve this...

I've been working my way through the first volume and knowing a bit of algebra has helped immensely. I recommend checking out volume 1 and see where it takes you!