Also, many more scientific concepts such as gradient descent are most easily expressed using the appropriate mathematic operators like the nabla operator for gradients.
I try to stay mindful of the limits of short term brain storage.. if there's over 7 unique variables or interval indicators, it might be tough for someone unfamiliar with what's going on to pick it up without wasting extra time noodling on it. When you've got a sum over 'm', a sum over 'm,n', a product sum over 'n,x' and another over 'x', etc.. sometimes it's just easier to bite the bullet and give a visual aid so the reader's not trying to envision all these overlapping regions in his head while still trying to remember what each is supposed to represent in real life.