0

I've checked out a few answers about multiplying large matrices, but none have seemed to articulate what I am looking for.

My data is a sparse 25000x4096 matrix, let's call it D.

I'm looking to perform D*D', but it's ungodly slow even with sparse data types. I feel like there must be some more efficiently way because the result is symmetric anyway, so half the operations are redundant.

Is there an operation I'm missing?

Community
  • 1
  • 1
marcman
  • 3,233
  • 4
  • 36
  • 71

1 Answers1

0

Performance of such algorithms depends a lot from memory schemes that you use for storing the matrix. For example, if your algorithm operates with linked lists, its performance may be significantly impaired in comparison with using something like an adjacency structure (that term comes from the theory of graphs).