next up previous print clean
Next: Acknowledgments Up: M. Clapp: The oddities Previous: Results

Conclusions

I have shown that the regularization parameter $\epsilon$ and the number of iterations niter can have a significant effect on the result of regularized least-squares inversion. The choice of niter generally depends on extraneous concerns such as available computer power and the best choice for $\epsilon$ depends on the choice of niter. For the experiments carried out in this paper, the best combination was niter=3 and $\epsilon=1$, but it is not likely that this is the ``ideal'' combination even for this problem, let alone other other inversion problems.
next up previous print clean
Next: Acknowledgments Up: M. Clapp: The oddities Previous: Results
Stanford Exploration Project
7/8/2003