Leonard Ritter on Nostr: speaking of spiraling towards results, is there a technique for exploiting ...

speaking of spiraling towards results, is there a technique for exploiting momentum-based gradient descent situations where we recognize that we are "spiraling"?

namely, what's the mathematical basis for recognizing and then extrapolating spiraling motion to find the center point?

i'd probably try something like fitting slices of path history to ellipses and then average their center...