Why does Relative Robust Representations (RRR) of Matrices determine eigenvalues to high relative accuracy?

47 Views Asked by At

I am trying to understand the MRRR algorithm for finding eigenpairs. As part of this I am reading Parletts and Dhillons paper on "Relatively robust representations of symmetric tridiagonals". I am having a hard time understanding the mathematical details (I am not a mathematician) as to why this representation is superior for achieving high relative accuracy. Would anyone here be able to explain it, or perhaps provide me with some additional literature on the subject?