[Insight-users] On multipleValuedcost function and multipleValuedOptimizers

[Markus Weigert]

Dear all,

I am dealing with a few basic questions on the usage of multiple valued
costfunctions and multiple valued optimizers.
As far as I understand, multiple valued costfunctions are functions,
which return not one but a couple of values for a given parameterset.
So, one could imagine to construct a multiple (two) valued costfunction on 
basis of a image intensity function and a point based distance function.

First:
How do multiple valued optimizers optimize such a cost function (do they 
optimize a weighted mean or  something similar of the values returned by the 
costfunction or do they really treat all values for itself and then: how?)

Second:
What about optimizers needing gradients?
Are gradients also implemented for multiple valued costfunctions and are there
optimizers available, which actually use the gradients.


Kind regards,
Markus Weigert