Next: Estimate of Marquardt
Up: 4: Minimization
Previous: Why is M
- Marquardt's method is a simpler technique but with similar goals
to the eigenvalue method.
- One replaces

- where I is the identity matrix.
- So this means we replace M by
- Clearly, if
, then the two formalism's are
roughly equivalent.
- Eigenvectors with eigenvalues >Q, are shifted by naive
prediction while those with eigenvalue <<Q, have their shift
drastically reduced.
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu