I suggest you to not learn with books. Look for something you want to do with your neural net (in my case I started with face recognition in pictures and am now working on a self driving car)
You should learn the math behind it first and stuff, then over to optimization and so on.
I can really recommend you 'sentdex' who created some really interesting things.
Also, once you have a bit of knowledge go look at the structures of different neural networks from different companies (google, facebook, nvidia), how many layers they use, how many neurons per layer, which optimizer bla bla bla.
Apart from that, I recommend you, you look for something you'd like to create and then google a net for it. There will nearly always be a neural net for it which you can modify understand and thus learn doing them yourself.
If you haven't decided on a framework I strongly recommend you TensorFlow or Keras
If you have any more explicit questions you may ask me in PM if you want.