[fixed] Sketcher Levenberg-Marquardt solver precision issues?

Have some feature requests, feedback, cool stuff to share, or want to know where FreeCAD is going? This is the place.
Forum rules
Be nice to others! Read the FreeCAD code of conduct!
triplus
Posts: 9475
Joined: Mon Dec 12, 2011 4:45 pm

Re: Sketcher Levenberg-Marquardt solver precision issues? [now with steps to reproduce!]

Postby triplus » Fri Mar 27, 2020 6:51 pm

DeepSOIC wrote:
Tue Mar 24, 2020 1:06 am
PR #3251
I can confirm the issue reported in this thread is now fixed.
triplus wrote:
Mon Mar 23, 2020 12:19 am
Not a silver bullet, though. Increasing the length to lets say 400 and going back to 40 starts causing "flip situation" for LevenbergMarquardt.
Currently Tau setting for LevenbergMarquardt is by default set to 0.001. I have set it to 1e-10 and in the above examples LevenbergMarquardt can handle jumps from 10m to 40mm without introducing the flip, no performance penalty observed.
chrisb
Posts: 28117
Joined: Tue Mar 17, 2015 9:14 am

Re: [fixed] Sketcher Levenberg-Marquardt solver precision issues?

Postby chrisb » Fri Mar 27, 2020 7:15 pm

I have the feeling that 0.19 will become a high quality version with many longterm issues fixed.
Thanks!
User avatar
Kunda1
Posts: 8461
Joined: Thu Jan 05, 2017 9:03 pm

Re: [fixed] Sketcher Levenberg-Marquardt solver precision issues?

Postby Kunda1 » Sun Mar 29, 2020 12:16 am

I remember there being (a) ticket(s) in the bugtracker related to "Levenberg-Marquardt solver"... let me see if I can find it..?

Edit:
issue #3348
issue #3773
issue #3961
issue #3286
Want to contribute back to FC? Checkout:
#lowhangingfruit | Use the Source, Luke. | How to Help FreeCAD | How to report FC bugs and features