|
NifTK
16.4.1 - 0798f20
CMIC's Translational Medical Imaging Platform
|
Implements Powell optimization using Brent line search. More...
Implements Powell optimization using Brent line search.
The code in this class was adapted from the Wikipedia and the netlib.org zeroin function.
http://www.netlib.org/go/zeroin.f http://en.wikipedia.org/wiki/Brent_method http://en.wikipedia.org/wiki/Golden_section_search
This optimizer needs a cost function. Partial derivatives of that function are not required.
For an N-dimensional parameter space, each iteration minimizes(maximizes) the function in N (initially orthogonal) directions. Typically only 2-5 iterations are required. If gradients are available, consider a conjugate gradient line search strategy.
The SetStepLength determines the initial distance to step in a line direction when bounding the minimum (using bracketing triple spaced using a golden search strategy).
The StepTolerance terminates optimization when the parameter values are known to be within this (scaled) distance of the local extreme.
The ValueTolerance terminates optimization when the cost function values at the current parameters and at the local extreme are likely (within a second order approximation) to be within this is tolerance.
1.8.8