Books are more systematic and can shine light in understanding the whole picture, and also the detail. Good books are also self-contained, so you won't encounter the problems while reading a paper, e.g an unknown term, a hard-understanding formulation, etc.
But books are also prone to become obsolete very quickly, so we should also cover papers to keep us to the edge.
Some of books I have read can be found here
We might not have the chance to study in some prestigious universities, but we can attend their classes now(thanks to MOOC). But we always feel overwhelmed with so many courses, so I give my list of courses and review here.
Besides theory, we also need to put things in practice, since most of us are not supposed to be a scientist(or researcher), but to be a engineer to convert the theory into the real world application. Therefore, learning from tutorials is always helpful and also interesting(AI is like magic in some ways 😀).
ML/DL as a fast-evolving field, it's always best to catch the latest progress from reading the good papers.
So I have my own papers reading.
As a enginner(programmer/coder, or whatever you name it), we all know "talk is cheap show me the codes", so we need to put the theory into application, which to make people an easier life, show our skill, or just for fun.
I have several ML related projects, which might be toys, but it must have fun in its own.