Hacker News Reader

Are you guys able to offer any guidance on numerical stability of linear algebra problems? For example, the Woodbury matrix identity's stability seems uncertain based on https://mathoverflow.net/questions/80340/special-considerati...

I think it depends on the nature of the matrices; it doesn’t make sense to use if your A matrix is just some general dense matrix for example.

Why doesn't it make sense to use in that case?

The dimensions of (A-UBU’) are the same as the dimensions of A. After you apply Woodbury, you have

inv(A) - (other stuff)

So, it only makes sense if A is much easier to solve than (A-UBU’).