next up previous contents index
Next: What Operations Can One Up: Singular Value Decomposition Previous: Direct Methods   Contents   Index

Iterative Algorithms
  J. Demmel

Now we discuss the pros and cons of the methods for Hermitian eigenproblems from Chapter 4 as applied to $A^* A$, $AA^*$, or $H(A)$. We note that the special structure of $H(A)$ may be exploited to make Lanczos more efficient, as described in §6.3.3 below.

For simplicity, we let $B$ denote any one of the above three Hermitian matrices in the discussion below. The basic tradeoffs from Table 4.1 remain true; the choice of method depends on the following:

What differs from Chapter 4 is how the above decisions depend on the choice of $B$.



Subsections
next up previous contents index
Next: What Operations Can One Up: Singular Value Decomposition Previous: Direct Methods   Contents   Index
Susan Blackford 2000-11-20