[Insight-users] 8D Point Set Gradient

[Kevin H. Hobbs]

I would like some advice about producing the gradient at each point from
an eight dimensional set of points with an error value at each point.

I do not believe there is any ITK point set filter that does this, so I
was thinking that I would use the kd tree to find a simplex of points
for each point and then just average the change in positions and error.

There probably exists a better way to do that but the only things I've
found are for lower dimension surfaces.

For background, the eight dimensions are the maximal conductances in a
neuron model, and the error values are the difference of the voltage
output of the models and a known target model.

I want to compare the space around the target model under two
stimulation conditions and get some sense of how jagged the error
landscape is. This will be useful for choosing the stimulation to use in
fitting the model to an unknown real target neuron.

This feels like a watershed segmentation problem to me. The best
stimulation will produce a single watershed that empties at the target,
and worse stimulations will have many small basins.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 252 bytes
Desc: OpenPGP digital signature
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100122/dcc982f2/attachment.pgp>