Vectorizing matrix fitting to a curve

Hello,

I have three volume matrices A, B, C of 128x128x128 each.

A represents the data at t1 B represents the data at t2 C represents the data at t3

Now, I need to fit each corresponding voxel to a curve, so I'm using the fit function available in matlab

coefficients=fit(xdata,ydata,myfunction,options);

where

xdata=[t1,t2,t3]

and

ydata=[A(i,j,k),B(i,j,k),C(i,j,k)];

So the way I'm doing this right now is I have to loop through the 128x128x128 voxels of my matrices in order to obtain the fitting coefficients of the curve for each corresponding voxel at the different times and storing them in a cell:

for i=1:128
   for j=1:128
      for k=1:128
        coefficients{i*j*k}=fit([t1,t2,t3],[A(i,j,k),B(i,j,k),C(i,j,k)],myfunction,options);
end
end
end

This is of course taking a very long time.

Is there a way of vectorizing this or at least making it more computational efficient?

I appreciate any help.

Thank you

Answer 1

[HG]

Hi, I have an almost identical problem to the one described above. Did you find a solution to your problem?

Answer 2

[Dylan Marques]

Hi,

I am not aware of an easy solution for your question. A possible way of making it faster is to use parfor (https://uk.mathworks.com/help/parallel-computing/parfor.html) so that the for runs in multiple cores.

The faster solution is to make a vectorized version of fit. However this can be complicated to do and time consuming (your time, not the computer time). Which function are you trying to fit to your data?

Many thanks