Here, we assumed that u was a unit vector. Aren't there times where we would want to relax that condition?
Does there exist another formula of directional derivative other than the limit one?
I'm not sure if this was covered elsewhere and I missed it, but how do we handle the directional derivative if the function is not differentiable? Or do we avoid computing it altogether?
What are the common use cases of directional derivative in CG?
Similar to above question: how often are our datasets/objects 'smooth'? Is much effort spent on ensuring models are?