Objective function is returning Inf or NaN values at initial point. lsqcurvefit cannot continue.

Question

Nhat Nguyen 2023-10-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2032339-objective-function-is-returning-inf-or-nan-values-at-initial-point-lsqcurvefit-cannot-continue

编辑： Torsten 2023-10-12

Hi guys, I tried to fitting my data with following code

k0= 3.5e+05; % (s-1)
lamda=7.5e-10; % (m)
n=0.92e-20;
KbT=4.11e-21; % (J)
p=[k0, lamda, n];
func=@(p,ydata)2*p(1)*p(2)*sinh((0.072*cosd(70)-0.0623.*cosd(ydata))/(2*KbT*p(3)));
lb=[1e+1; 1e-12; 1e-15]; % lower bound
ub=[1e+9; 1e-6; 1e-25]; % upper bound
f=lsqcurvefit(func, p, ydata, xdata, lb,ub);
plot(xdata,ydata,'ko',xdata,func(f,ydata),'b-')

But, when I ran, the code results in:

Error using lsqncommon
Objective function is returning Inf or NaN values at initial point. lsqcurvefit cannot continue.
Error in lsqcurvefit (line 274)
    lsqncommon(funfcn,xCurrent,lb,ub,options,defaultopt,optimgetFlag,caller,...
Error in mkt (line 21)
f=lsqcurvefit(func, p, ydata, xdata, lb,ub);

I tried to look at my data, but it looks find to me. Please see the attachment for the data reference, column 1 is xdata, column 2 is ydata.

Thanks for suggestion.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Torsten 2023-10-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2032339-objective-function-is-returning-inf-or-nan-values-at-initial-point-lsqcurvefit-cannot-continue#answer_1331689

编辑：Torsten 2023-10-12

在 MATLAB Online 中打开

Before you call lsqcurvefit, you should call your function "func" with the initial guess values for the parameters to see what it returns. So insert the command

test = func(p,ydata)

before the line

f=lsqcurvefit(func, p, ydata, xdata, lb,ub);

Further it is usual that you have a function ydata = func(p,xdata). You reversed the roles of xdata and ydata: xdata = func(p,ydata). You should check whether this is really what you want.

Further your model depends on only one parameter, not three. The expression

2*p(1)*p(2)/(2*KbT*p(3))

can be seen as one parameter P you try to estimate. It's not possible to distinguish between p(1), p(2) and p(3) because they are subsummed in one expression.

To understand this, imagine you try to fit a model

ydata = p(1)*p(2)*xdata

If the optimizer returns p(1) = 2, p(2) = 5, would this in any way be better than p(1) = 1, p(2) = 10 or p(1) = 2.5, p(2) = 4 ?