Timeline for Fitting a spherical surface given points with known distances from the (unknown) surface
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 9, 2024 at 20:00 | comment | added | lightxbulb | If you are talking about the $\frac{c-p_i}{\|c-p_i\|}$ term making the gradient "undefined", I think you should evaluate that to zero if the denominator is zero. That is, if you get a NaN turn it into a 0. As far as gradient descent goes I suppose that it should work, but you may have to do a line search for the step size. But if you are unhappy with it you can look at conjugate gradients, damped Newton or Gauss-Newton/Levenberg-Marquardt. I wanted to make a python code to illustrate that but I just didn't find any time, sorry. | |
| S Dec 9, 2024 at 4:54 | review | First answers | |||
| Dec 15, 2024 at 4:02 | |||||
| S Dec 9, 2024 at 4:54 | history | answered | Norman | CC BY-SA 4.0 |