>you can start with concrete problems and generalize from there
You could. But that isn't the only way. You should really read up on the way math is( was ?) taught in gottingen & eastern europe & romania & russia. Speaking of which, I learnt most of the machinery behind matrices in India in high school, with no reference to where it came from or what its applications were. I could compute the determinant of square matrices, compute adjoints, inverses, multiply two matrices....all of this without knowing wtf a matrix was or what its real life use might be. This is true of most high school students in India, simply because of the way the curriculum is structured. Its only when I took at a graduate level computer graphics course in my late 20s did I see that linear transforms were actually matrices, so you could do rotations and translations of vectors by matrix multiplication. So real life applications don't have to precede abstract concepts - you can make a hell a lot of progress by approaching stuff the other way around.
> I could compute the determinant of square matrices,
I learned to do this too---and then promptly forgot---several times over until someone finally clued me in to the fact that the determinant is the scale factor of the linear transformation associated with the matrix. Then everything clicked.
It's not just that I couldn't remember how to define/calculate the determinant, I also couldn't apply it usefully in my (proof-based) linear algebra class because I didn't have any intuition for it or any way to connect it to other concepts.
You could. But that isn't the only way. You should really read up on the way math is( was ?) taught in gottingen & eastern europe & romania & russia. Speaking of which, I learnt most of the machinery behind matrices in India in high school, with no reference to where it came from or what its applications were. I could compute the determinant of square matrices, compute adjoints, inverses, multiply two matrices....all of this without knowing wtf a matrix was or what its real life use might be. This is true of most high school students in India, simply because of the way the curriculum is structured. Its only when I took at a graduate level computer graphics course in my late 20s did I see that linear transforms were actually matrices, so you could do rotations and translations of vectors by matrix multiplication. So real life applications don't have to precede abstract concepts - you can make a hell a lot of progress by approaching stuff the other way around.