I have a gigantic matrix whose svd needs to be computed. However the matrix actually is made of rows and columns with repetitions.
The matrix of unique rows and columns, $M_{ij}$ is small enough to be easily handled. The row $i$ repeats $R_i$ times and column $j$ repeats $C_j$ times.
Given (M,R,C) is it possible to reconstruct the singular values of the full matrix from the singular values of $M$ alone ?