next up previous print clean
Next: Implementation Up: R. Clapp: Regularization Previous: Results


Instead of approximating the Hessian with a diagonal matrix we can attempt to estimate the least squares inverse using a conjugate gradient solver. The model is preconditioned by using polynomial division to apply the helical derivative and the new preconditioned variable $\bf p$ is estimated through  
Q(\bf p) = \vert\vert\bf d- \bf L \bf C \bf p\vert\vert^2 + \epsilon^2 \vert\vert \bf p\vert\vert^2
,\end{displaymath} (11)
where $\bf m= \bf C \bf p$.