| forum.alglib.net http://forum.alglib.net/ |
|
| Trying to understand alglib.minlmoptimize http://forum.alglib.net/viewtopic.php?f=2&t=581 |
Page 1 of 1 |
| Author: | jkh13 [ Thu May 24, 2012 5:58 pm ] |
| Post subject: | Trying to understand alglib.minlmoptimize |
I am using minlmoptimise() with minlmcreatev() and after 12 iterations it finished the optimisation. However it didn't get great results, I checked the report object (alglib.minlmresults(state, out unknowns, out rep);) and it reported 12 iterations with the termination code 2 - Sigma threshold reached. When debugging I checked the fi object inside the state object and summed up all the fi differences, these summed up to a value >100 which is far larger than my sigma threshold (double epsg = 0.0000000001;). I was wondering how it came to this conclusion? Any help would be greatly appreciated. Thanks |
|
| Author: | Sergey.Bochkanov [ Fri May 25, 2012 11:02 am ] | ||
| Post subject: | Re: Trying to understand alglib.minlmoptimize | ||
It is possible that you have so called "bad fit problem" (see below). In this case, you can have quite large function value at the solution (where gradient becomes zero). If you don't think that you have "bad fit" situation, you can post your function, so I will be able to study algorithm behavior on your problem.
|
|||
| Page 1 of 1 | All times are UTC |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|