How to speed up or parallelise multiple calls to lsqcurvefit

I have some code that has to run lsqcurvefit multiple times on different datasets. Typically I have around 20,000-50,000 datasets that all require independant fitting. Is there any way to speed the process up by e.g. splitting the work across multiple threads?

I've tried the basic tools in the parallelisation toolbox but did not notice any speed difference.

[Matt J]

Is there any way to speed the process up by e.g. splitting the work across multiple threads?

Are you sure you have any unused threads available? If you are on a Windows OS, show a screen shot of the Task Manager so that we can see the utilization of the cores.

[Nathan Bush]

Across an 8 core machine I have an average utilisation of around 25%, so i figure I could conceivably get a factor of x2 speed up if I could somehow achieve better CPU utilisation.

[Matt J]

And how does that utiltization change when you use Parallel Computing Toolbox? And how were you using the toolbox? With parfor?

[Nathan Bush]

Thanks for the help. The utilisation was pretty much the same, an the execution time was only a few % better. Yes - it was a simple parfor execution, and I didn't change any of the settings. I checked in preferences that Matlab was using the correct number of workers (8) and it seemed ok.

Answer 1

[Matt J]

This thread may help,

https://www.mathworks.com/matlabcentral/answers/431697-make-curve-fitting-faster