I ended up giving up and not bothering with deep learning and neural nets until the amazing, legendary, awesome CS231n was made available taught by Andrej Karpathy.
I don't know what it is about the way Andrej teaches but it's so good that I literally binge watched all of CS231n over a weekend and then went back to the start and watched again while working through the exercises.
Just goes to show that not every teaching style will fit every student and that's why online learning has a big future - teaching/learning style compatibility is a hugely under-appreciated issue.
I admire Geoff Hinton. I'm also proud of his accomplishments as a fellow Canadian. But I'm not in sync with the way he explains stuff though I'll always watch every lecture and presentation he puts out on video.
Random anecdote; I learned how to solve Rubik’s cube and get to sub 30 seconds by watching his YouTube channel ~10 years ago (He was a PhD student at Stanford at the time). I was always amazed by his teaching skills even back then.
Karpathy's CS231n was indeed another great course. We're really spoiled for choice these days.
Two years ago Geoff Hinton requested his course be removed because it was out of date.
At a university, at least at a post-graduate level, you are probably going to be taught by a deep learning researcher passionate about the subject, with the added advantage of support if you struggle with some concepts.