If you're self-learning, you need to pick what you're going to learn from. Which means, ultimately, you need to pick what you will learn. If you don't know what it is (or what it's called), that can be very difficult.
Anecdote: I have zero background in graphics, but found myself needing to do some image processing. I spent about a couple days learning what I could about the problem space and deriving then writing an algorithm. It worked pretty well and I felt very clever about myself. A few days later I ended up googling part of the algorithm I wrote to try to better understand some edge cases. I ended up on Wikipedia staring at the algorithm I had written, only better, and simpler, and faster, and followed by a few paragraphs entitled "how to derive this algorithm from first principles". It was amazing, but I felt decidedly less clever about myself.
People say that in the age of the Internet, being taught this stuff isn't terribly important because you can just look things up. A caveat is you need to know what something is called to look it up. I've come to the opinion that a big portion of the value of higher education comes from being familiarized with solved problems and their names by someone who (hopefully) knows what you'll most likely want to use at some point.