Reduce the computional time to calculate the mutilplying two matrixes

1 visualización (últimos 30 días)
Thu Nguyen
Thu Nguyen el 17 de Mayo de 2019
Comentada: Thu Nguyen el 17 de Mayo de 2019
I calculate this loop for thousands of time.
How can i optimize it to reduce the computational time?
for i=1:n
f(i,:)=w'*Q0(:,:,i)
end
with w =(2000,1) matrix and Q0=(2000,2000,100)
Thanks a lot.
  2 comentarios
dpb
dpb el 17 de Mayo de 2019
  1. Preallocate f -- f=zeros(n,numel(w));
  2. Reorient w outside the loop
The first may help enough to be noticeable if haven't; the second is very minor aid to the optimizer and likely will make little, if any, difference.
Thu Nguyen
Thu Nguyen el 17 de Mayo de 2019
Thank you a lot. I also made preallocate and reorient as you made and only reduce little time. I need reduce more such as ~ 100 times.

Iniciar sesión para comentar.

Respuestas (1)

David Goodmanson
David Goodmanson el 17 de Mayo de 2019
Editada: David Goodmanson el 17 de Mayo de 2019
Hi Thu,
try
% N = 1000;
% M = 200;
Q00 = reshape(Q0,N,N*M);
ff = reshape(w'*Q00,N,M)';
This is approximately four times faster on my PC, compared to the first way with f preallocated.
  1 comentario
Thu Nguyen
Thu Nguyen el 17 de Mayo de 2019
Thank you so much. I found this solution very interesting. But on my PC, it reduce time a little. I want to reduce the computational time about 100 times. Any solution?

Iniciar sesión para comentar.

Categorías

Más información sobre Robust Control Toolbox en Help Center y File Exchange.

Community Treasure Hunt

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

Start Hunting!

Translated by