Least Square Complex Non-Linear Curve Fitting

I am looking for Least Square Curve fitting of non linear complex numbers. I have this objective functions.
F = (a + bD) * Y^2 + (c + dD) * Z^2
Where F, a, b, Y, c, d, Z are all complex numbers. D is real number. To add that inputs to the model are Y, Z and D. Unknown are a,b, c and d. F is the known output. I found some article but those are talking about Real inputs and outputs with complex numbered coefficients. However, my case has all complex numbers except D.

回答(1 个)

You won't be able to identify a,b,c,d,D separately.
Switch to the model
F = const1 * Y^2 + const2 * Z^2
and identify complex-valued const1 and const2 by splitting const1, const2, F, Y and Z in real and imaginary part.
Best wishes
Torsten.

3 个评论

Thanks for prompt comment. Sorry to add that D is known and will be given as input to the model. So inputs to the model are Y, Z and D. Unknown are a,b, c and D. F is the known output.
I have updated my question to clarify my point.
My objection still holds even if D is known.
A complex number z can be written as z=a+b*D in many, many ways, e.g. z=0.5*z+D*0.5*z/D or z=0.2*z+D*0.8*z/D or ...
Thus the numbers a and b cannot be estimated independently which makes a separate fitting of the form z=a+D*b useless.
Best wishes
Torsten.

请先登录,再进行评论。

类别

帮助中心File Exchange 中查找有关 Fit Postprocessing 的更多信息

提问:

2017-2-2

编辑:

2017-2-2

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by