About Using the MATLAB built-in function in fitting process

Hello, I'm using custom equations in fitting my data. The problem is that these equations contain one or more MATLAB built-in functions (e.g., expint, lambertw,..) and it takes so long (about 20 minutes) to converge. Is there any way so that I can decrease the accuracy of these built-in functions so that the convergence become faster? Or any idea to accelerate the fitting process?

Thanks.

Answer 1

[Walter Roberson]

expint() is available both numerically and in the Symbolic Toolbox. lambertw is available only in the symbolic toolbox, but a numeric version of it is available in the File Exchange

If you need to use the symbolic toolbox then you can decrease computation time by setting Digits to lower.

[Amr Sadek]

Thank you for your answer. I use these equations in a GUI. So I'm using the MATLAB (numeric version) not the Symbolic Toolbox.

[Walter Roberson]

How were you calling a numeric lambertw ? Were you already using the version from the File Exchange?

[Amr Sadek]

I'm using MATLAB 2014a. The Lambertw function is a built-in function. I don't use any File Exchange.

[Walter Roberson]

Please use

which -all lambertw

to see which LambertW implementation you are getting. The two that are documented are Symbolic Toolbox, MATLAB Interface and Symbolic Toolbox, MuPAD Engine

There are a few different implementations in the File Exchange, including the one by Cleve, linked above, and another by D.A. Barry at http://www.mathworks.com/matlabcentral/fileexchange/3644-real-values-of-the-lambert-w-function

[Amr Sadek]

I'm actually using lambertw(x) for the first (principal) branch and lambertw(-1,x) for the second branch in my codes. Is that what you mean?

[Walter Roberson]

What does

which -all lambertw

say?

[Amr Sadek]

 which -all lambertw
D:\...\MATLAB 2014a\toolbox\symbolic\symbolic\@single\lambertw.m  % single method
D:\...\MATLAB 2014a\toolbox\symbolic\symbolic\@double\lambertw.m  % double method
D:\...\MATLAB 2014a\toolbox\symbolic\symbolic\@sym\lambertw.m     % sym method