My own take is that the "theory of matrices" (not a standard usage but should be) is worth learning for some (sp if you're going into hardcore numerics like sta

My own take is that the "theory of matrices" (not a standard usage but should be) is worth learning for some (sp if you're going into hardcore numerics like stability of algorithms) but not for everyone. It's like studying calculus as the theory of smooth curves -- it robs you of some potential for growth. Also: some people just want "math for deep learning" and that's mostly baby calculus: the chain rule and gradient descent.