How to fastly construct a sparse matrix with each column equal to some sparse vector

yu wen

2019 9 26

0 个回答

3 次查看（30 天）

0 个投票

Dear community,

I have 2000 sparse arrays H(i), i=1,...2000, with each of size 100000*5, now I'd like to construct a new sparse matrix X of size 100000*2000 with its i-th column equal to the first column of H(i). The simple way is just a loop:

X = sparse(100000,2000);

for i=1:2000

X(:,i) = H(:,1);

end

However, Such an assignment is slow. Is there any better way? Thanks !

2 个评论
显示无隐藏无

Fabio Freschi 2019-9-26

编辑：Fabio Freschi 2019-9-26

Not clear: is H an array (you use H(i)) or a matrix (in the loop: H(:,1)). or maybe a cell array where each H{i} is a 100000*5 matrix?

Can you share the data to play with?

yu wen 2019-9-26

Sorry for the unclear statement. H is a sparse matrix of size 100000*5 in the loop and will be assigned by some_function. I want to extract the first column of H to construct a new sparse matrix X, that is, each column of X is the first column of matrix H.

So the simple code is:

X = sparse(100000,2000);

for i = 1:2000

H = some_function(..); % get a new sparse matrix H of size 100000*5

X(:,i) = H(:,1); % extract the first column of H and put it into the i-th column of X

end

It works, but the Code Analyzer in Matlab detects such an index pattern for a sparse matrix that could be slow. I don't know how to accelerate the assignment.

请先登录，再进行评论。

请先登录，再回答此问题。

Follow Question