fitting with the custom model without using CurveFittingToolbox
in [Matlab]
|
From: Oznur Mete on 5 Mar 2010 06:42 Dear All, I am trying to define a custom fit model and fit to the data. Here is the piece of my code: ====================================== s = fitoptions('Method','NonlinearLeastSquares',... 'Lower',[0.9,0.9],... 'Upper',[1.1,1.1], ... 'Startpoint',[1 1]); f = fittype('a*exp(b*sqrt(sind(x)))','coefficients',{'a','b'},'independent','x','options', s); ph1= ph1(:); ch1 = ch1(:); [c2,gof2] = fit(ph1,ch1,f); ===================================== I received the following error message: =================================== ??? Complex value computed by model function, fitting cannot continue. Try using or tightening upper and lower bounds on coefficients. Error in ==> fit at 448 errstr = handleerr('curvefit:fit:complexValueComputed',errmsg,suppresserr); =========================== I do not prefer to use curve fitting toolbox, although it is a nice tool, I need to include the fitting as a part of a long and iterating code. Therefore I prefer to use custom model fitting. Concerning the error message I have played with the upper and lower bounds. The error message was still there... Ideas will be highly appreciated. Cheers, Oznur
From: Oznur Mete on 5 Mar 2010 08:04 "Oznur Mete" <oznur.mete(a)cern.ch> wrote in message <hmqqmt$4v8$1(a)fred.mathworks.com>... > Dear All, > > I am trying to define a custom fit model and fit to the data. > > Here is the piece of my code: > ====================================== > s = fitoptions('Method','NonlinearLeastSquares',... > 'Lower',[0.9,0.9],... > 'Upper',[1.1,1.1], ... > 'Startpoint',[1 1]); > f = fittype('a*exp(b*sqrt(sind(x)))','coefficients',{'a','b'},'independent','x','options', s); > ph1= ph1(:); > ch1 = ch1(:); > [c2,gof2] = fit(ph1,ch1,f); > ===================================== > > > I received the following error message: > =================================== > ??? Complex value computed by model function, fitting cannot continue. > Try using or tightening upper and lower bounds on coefficients. > > Error in ==> fit at 448 > errstr = handleerr('curvefit:fit:complexValueComputed',errmsg,suppresserr); > =========================== > > > I do not prefer to use curve fitting toolbox, although it is a nice tool, I need to include the fitting as a part of a long and iterating code. Therefore I prefer to use custom model fitting. > > Concerning the error message I have played with the upper and lower bounds. The error message was still there... > > Ideas will be highly appreciated. > > Cheers, > Oznur Typo: The code will be like this surely (with 'options'): ====================================== > s = fitoptions('Method','NonlinearLeastSquares',... > 'Lower',[0.9,0.9],... > 'Upper',[1.1,1.1], ... > 'Startpoint',[1 1]); > f = fittype('a*exp(b*sqrt(sind(x)))','coefficients',{'a','b'},'independent','x','options', s); > ph1= ph1(:); > ch1 = ch1(:); > [c2,gof2] = fit(ph1,ch1,f,'options',s); > ===================================== Oznur
From: Steven Lord on 5 Mar 2010 09:18 "Oznur Mete" <oznur.mete(a)cern.ch> wrote in message news:hmqqmt$4v8$1(a)fred.mathworks.com... > Dear All, > > I am trying to define a custom fit model and fit to the data. > > Here is the piece of my code: > ====================================== > s = fitoptions('Method','NonlinearLeastSquares',... > 'Lower',[0.9,0.9],... > 'Upper',[1.1,1.1], ... > 'Startpoint',[1 1]); > f = > fittype('a*exp(b*sqrt(sind(x)))','coefficients',{'a','b'},'independent','x','options', > s); Note that if x contains a value whose sine (when it's treated as an angle in degrees) is negative, your function will return a complex value. That's what the error message complained about. > ph1= ph1(:); > ch1 = ch1(:); > [c2,gof2] = fit(ph1,ch1,f); > ===================================== > > > I received the following error message: > =================================== > ??? Complex value computed by model function, fitting cannot continue. > Try using or tightening upper and lower bounds on coefficients. > > Error in ==> fit at 448 > errstr = > handleerr('curvefit:fit:complexValueComputed',errmsg,suppresserr); > =========================== > > > I do not prefer to use curve fitting toolbox, although it is a nice tool, > I need to include the fitting as a part of a long and iterating code. > Therefore I prefer to use custom model fitting. Well, the functions FITOPTIONS, FITTYPE, and FIT are all part of Curve Fitting Toolbox ... what I think you mean is that you don't want to use the Curve Fitting Toolbox GUI CFTOOL, right? If you're not aware of this feature, I'd like to let you know you could use the CFTOOL GUI to experiment with getting the fit to look and act the way you wanted it, then generate code from the GUI for use in your "long and iterating code". http://www.mathworks.com/access/helpdesk/help/toolbox/curvefit/bqxoya6.html > Concerning the error message I have played with the upper and lower > bounds. The error message was still there... In some cases, adjusting the bounds on the parameters could avoid the function returning a complex value. f = fittype('sqrt(a*x)', 'coefficients', 'a', 'independent', 'x') Assuming all your x values are nonnegative, if your original lower bound on a included negative numbers, changing the lower bound to be at least 0 would resolve the "Complex value computed by model function" issue. But in this case, it won't. > Ideas will be highly appreciated. Change the form of your function so that it doesn't try to compute the square root of a negative value or change your phi data so it doesn't contain any values whose sind is negative. -- Steve Lord slord(a)mathworks.com comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ
From: Oznur Mete on 5 Mar 2010 09:46 "Steven Lord" <slord(a)mathworks.com> wrote in message <hmr3rs$ctv$1(a)fred.mathworks.com>... > > "Oznur Mete" <oznur.mete(a)cern.ch> wrote in message > news:hmqqmt$4v8$1(a)fred.mathworks.com... > > Dear All, > > > > I am trying to define a custom fit model and fit to the data. > > > > Here is the piece of my code: > > ====================================== > > s = fitoptions('Method','NonlinearLeastSquares',... > > 'Lower',[0.9,0.9],... > > 'Upper',[1.1,1.1], ... > > 'Startpoint',[1 1]); > > f = > > fittype('a*exp(b*sqrt(sind(x)))','coefficients',{'a','b'},'independent','x','options', > > s); > > Note that if x contains a value whose sine (when it's treated as an angle in > degrees) is negative, your function will return a complex value. That's > what the error message complained about. > > > ph1= ph1(:); > > ch1 = ch1(:); > > [c2,gof2] = fit(ph1,ch1,f); > > ===================================== > > > > > > I received the following error message: > > =================================== > > ??? Complex value computed by model function, fitting cannot continue. > > Try using or tightening upper and lower bounds on coefficients. > > > > Error in ==> fit at 448 > > errstr = > > handleerr('curvefit:fit:complexValueComputed',errmsg,suppresserr); > > =========================== > > > > > > I do not prefer to use curve fitting toolbox, although it is a nice tool, > > I need to include the fitting as a part of a long and iterating code. > > Therefore I prefer to use custom model fitting. > > Well, the functions FITOPTIONS, FITTYPE, and FIT are all part of Curve > Fitting Toolbox ... what I think you mean is that you don't want to use the > Curve Fitting Toolbox GUI CFTOOL, right? > > If you're not aware of this feature, I'd like to let you know you could use > the CFTOOL GUI to experiment with getting the fit to look and act the way > you wanted it, then generate code from the GUI for use in your "long and > iterating code". > > http://www.mathworks.com/access/helpdesk/help/toolbox/curvefit/bqxoya6.html > > > Concerning the error message I have played with the upper and lower > > bounds. The error message was still there... > > In some cases, adjusting the bounds on the parameters could avoid the > function returning a complex value. > > f = fittype('sqrt(a*x)', 'coefficients', 'a', 'independent', 'x') > > Assuming all your x values are nonnegative, if your original lower bound on > a included negative numbers, changing the lower bound to be at least 0 would > resolve the "Complex value computed by model function" issue. But in this > case, it won't. > > > Ideas will be highly appreciated. > > Change the form of your function so that it doesn't try to compute the > square root of a negative value or change your phi data so it doesn't > contain any values whose sind is negative. > > -- > Steve Lord > slord(a)mathworks.com > comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ > Thanks Steve! Yes, I have just realized that the error's gone when I changed my x-values, due to the reason you explained above. Yes I did not want to use the GUI but seems I can go to this way and extract the code. Cheers, Oznur
|
Pages: 1 Prev: load matlab file (or create if does not exist) Next: continue m-file after using impoint |