There are two matrix A and B. The dimension of A and B are both 100000 by 100000. Matrix B is a symmetric matrix. I would like to calculate ABA. I just use the julia code
A*B*A
to do that. But it will take a long time. I am just wondering if there is any better ways to do it.
If they’re 32-bit dense matrices, those are 40 gigabytes each. Unless they have some kind of special structure like sparsity or that they’re a low rank product, it will be hard to do much better than just multiplying them directly. If they are sparse or low rank somehow, then you can probably take advantage of that somehow but we don't have enough information.