Sparse matrix operations cause extra non-zero values (very small)

I have two big sparse matrix A, B and I want to get

C=A\B

However, many of the elements in C are actually very small (<1e-10), which I think is due to precision. This would lead to a significant increase of the memory to store C.

My question is, what should I do to eliminate those small numbers, before actually storing them? I don't want to make some selections to the elements of C after its creation, which already takes up memories.

Answer 1

[Aritra]

Hi,

As per my understanding you are trying to convert values under a certain threshold (<1e-10) to 0.

You can make use of logical indexing for the task. Suppose you have a vector A with 5 elements, and you want to convert the values less than 3 to 0. Consider the following example illustrating the idea:

A = [1,2,3,4,5];
t = A(1:end)<3;
A(t) = 0;  

For more details on logical indexing, you can refer to the below documentation on MATRIX Indexing in MATLAB:

https://in.mathworks.com/company/newsletters/articles/matrix-indexing-in-matlab.html