Next: Outlook
Up: Claerbout: Anti-crosstalk
Previous: Crosstalk in a more
For warm up we linearize in the simplest possible way.
Suppose we allow only
to vary keeping
fixed.
We put
on the diagonal of a matrix, say
.
The regression for anti-crosstalk is now
Define the element-by-element cross product of
times
to be
.
Now let us linearize the full non-linear
anti-crosstalk regularization.
Let a single element of
be decomposed as a base plus a perturbation
.
A single component of the vector
is
.
Linearizing the product (neglecting the product of the perturbations) gives
|
(8) |
This is one component. We seek an expression for all.
It will be a vector which is a product of a matrix with a vector.
We want no unknowns in matrices; we want them all in vectors
so we will know how to solve for them.
|
(9) |
Express the perturbation parts of the vectors as functions
of the model space
|
(10) |
This vector should be viewed under many windows (triangle shaped, for example).
Under each window we hope to see the product have a small value.
The desired anti-crosstalk regression
is to minimize the length of the vector below by variation
of the model parameters
and
.
|
(11) |
This regression augments our usual regularizations.
Perhaps it partially or significantly supplants them.
Unfortunately, it requires yet another epsilon.
Upon finding
and
we update the base model
and iterate.
Next: Outlook
Up: Claerbout: Anti-crosstalk
Previous: Crosstalk in a more
2008-10-28