Just a bit more detail on why this problem is particularly important in modern ML compared to physics or linear algebra, and also mention that we don't only address the need to remember different axes, but also to define operations that act on different subsets of the axes and propagate the rest.
Just a bit more detail on why this problem is particularly important in modern ML compared to physics or linear algebra, and also mention that we don't only address the need to remember different axes, but also to define operations that act on different subsets of the axes and propagate the rest.