![]() Registered Member ![]()
|
Hello
I have some questions about non-linear optimization with our Unsupported/LevenbergMarquardt MINPACK implementation: - Parameter boundaries In some MINPACK implementations (mpfit, alglib) it is possible to set upper/lower boundaries to unknowns. It is possible in Eigen? - Weighted non-linear least-squares I'd like to set a weighting factor to equations from 1 to 0 so that equations with higher factor would weight more than others. How could it be implemented? Best regards Iñaki |
![]() Registered Member ![]()
|
Progress on the topics during this weekend:
- Parameter boundaries -- To force positive solutions in val: using abs(val) instead of val to calculate residual -- To force solutions in val over certain value min: : using abs(val - min) + min instead of val to calculate residual - Weighted non-linear least-squares Calculate for each value a weight so that residual = (y - f(x))*weight Best regards Iñaki |
Registered users: bartoloni, Bing [Bot], Google [Bot], Yahoo [Bot]