Rootfinding on manifold

[Tamas_Papp]

@dawbarton

Both of these approaches requires quite a bit of knowledge to make them work and so might not be appropriate for you.

I always appreciate being politely informed about my lack of knowledge But links to resources or examples would be welcome so that I could remedy this.

@dawbarton

replace the linear solve within the nonlinear root finder with a least-squares solution.

Unfortunately, this goes to wacky local minima about 20% of the time.

I am wondering whether there is a trust-region quasi-variant of Gauss-Newton [EDIT: maybe Levenberg–Marquardt?]