saving multiplications with zeros in a loop

%if true
for i iterations
Matrix = Matrix.*phase %Matrix has lot of zeros I want to skip multiplications with zero. Is it possible to do it in a single line without using if else or any loops
%end
Please dont use sparse, both matrix and phase are two-dimensional array

 Respuesta aceptada

Walter Roberson
Walter Roberson el 13 de Sept. de 2018
Mask = Matrix ~= 0;
for i iterations
Matrix(mask) = Matrix(mask) * phase ;
end
This assumes that phase is a scalar nonzero value. It does not apply if phase is a vector or 2d array.
If phase is a 2d array then
phases = sparse(phase) ;
Matrix = sparse(Matrix) ;
for i iterations
Matrix = Matrix * phases;
end
Or better yet,
Matrix = sparse(Matrix) * (phase^iterations) ;
with no loop.

4 comentarios

In your revised specifications of 2D matrices and using .* then
mask = Matrix ~= 0; Matrix(mask) = Matrix(mask) .* (phase(mask).^iterations);
This does more multiplications than are strictly necessary if entries in phase can be 0. This code uses two statements; optimizing the multiplications for phase potentially being 0 would require three statements.
Walter Roberson
Walter Roberson el 14 de Sept. de 2018
There are also optimizations available to reduce the multiplications cost of phase.^iterations but at the potential cost of more memory. These are easiest to code if iterations is a constant. The requirement to use only a single line of code acts against the requirement to reduce multiplications. As I indicated before, if efficiency is the top priority then you should give up on trying to force everything into one line.
D_coder
D_coder el 14 de Sept. de 2018
Editada: Walter Roberson el 14 de Sept. de 2018
actually this code is slow as compared to sparse. And sparse is slower as compared to the normal multiplication.Here is my detailed question. https://www.mathworks.com/matlabcentral/answers/418975-sparse-and-binary-mask
D_coder
D_coder el 14 de Sept. de 2018
Hey Walter my top priority is speed and saving multiplications.

Iniciar sesión para comentar.

Más respuestas (0)

Categorías

Más información sobre Quadratic Programming and Cone Programming en Centro de ayuda y File Exchange.

Etiquetas

Preguntada:

el 13 de Sept. de 2018

Comentada:

el 14 de Sept. de 2018

Community Treasure Hunt

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

Start Hunting!

Translated by