Let's say we have a huge matrix $A \in \mathbb{R}^{m \times n}$ where n is more than a million.
I would like to get a result of SVD as $U\Sigma V^{T} = A$.
Actually, the matrix $A$ is continuously changed slightly, for example, some of the rows are changed as follows $A[i, :] \leftarrow a^{T}_{i}$.
Calculating the whole matrices for the updated $A$ takes a lot and I think it is inefficient. I would like to know if there is any way to reuse the past SVD results for the new SVD or so.
Any comments would be appreciated.