Matlab arrays preallocation techniques

1 次查看（过去 30 天）

Giovanni Gardan 2020-5-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/532768-matlab-arrays-preallocation-techniques

评论： Stephen23 2020-5-25

I'm trying to make my code faster to run. In literature, one of technique is preallocating arrays used in the code.

Let's suppose we have a 1000x1000 matrix to pre-allocate before a while loop. It is known that sparse form save memory, but save also time?

Which of the following two options is better?

%First way
A = zeros(1000);
%Second way
A = sparse(1000);

3 个评论
显示 1更早的评论隐藏 1更早的评论

Giovanni Gardan 2020-5-25

Thank you for the answer.

Yes you are right, but seems like in some cases the first way is faster, in other cases the secondo way is faster. I'd like to know if there is a general rule to treat preallocation

Stephen23 2020-5-25

In general adding new data to sparse arrays is slow. Rather than preallocating the entire sparse array the usual efficient approach is to build up vectors of the indices and the data in the loop and after the loop convert them to sparse.

Your array is not very large I doubt that there is much point in defining it as sparse anyway.

请先登录，再进行评论。

请先登录，再回答此问题。