how to do SVD using the covariance matrix

355 Views Asked by At

I have an $(N \times M)$ matrix $A$ with $M \gg N$, $M$ being millions and $N$ hundreds, and I want to do $SVD$ on the matrix $A$. Can I do this calculation using $A\cdot A$ (the covariance matrix)?

1

There are 1 best solutions below

2
On

You cannot. Once you built $A^TA$ the left hand side singular vectors are lost. You can only solve it for the right vectors. Of course square root of eigenvalues will be your singular values.