1

I need to fit a model's parameters to observational data, but the model's precision is low, and it isn't smooth relative to the parameters. I.e. when a parameter is changed only so slightly, the output values may remain the same.

Due to this fact, lmfit runs the model for a few rounds, variating the parameters, and then gives up. How can I specify the minimum parameter variation step so the changes it makes are actually meaningful? I'm currently using the leastsq method.

dpq
  • 9,028
  • 10
  • 49
  • 69
  • is it not possible to use some sort of a `for` loop that loops through a list of your parameters for each of the initial guess? – Srivatsan Apr 06 '15 at 13:37
  • That's called "exhaustive search", and it's not a good option for big problems. This kind of software packages have been designed specially to optimize this task, and there's quite a bit of useful math involved to make things faster. So, no. – dpq Apr 06 '15 at 13:59
  • Possible duplicate: http://stackoverflow.com/questions/20478949/how-to-force-larger-steps-on-scipy-optimize-functions – dpq Apr 06 '15 at 14:36

0 Answers0