Plugin type: minimizer/singlecost
Prev Chapter 3. Plugin Reference 

Plugin type: minimizer/singlecost

These plug-ins provide optimizers of many-to-one functions

Plugins:

gdsq gsl nlopt

gdsq

Gradient descent with quadratic step estimation. Supported parameters are:

NameTypeDefaultDescription
ftolrdouble0Stop if the relative change of the criterion is below.
gtoladouble0Stop if the inf-norm of the gradient is below this value.
maxiteruint100Stopping criterion: the maximum number of iterations
scaledouble2Fallback fixed step size scaling
stepdouble0.1Initial step size
xtoladouble0Stop if the inf-norm of x-update is below this value.

gsl

optimizer plugin based on the multimin optimizers ofthe GNU Scientific Library (GSL) https://www.gnu.org/software/gsl/. Supported parameters are:

NameTypeDefaultDescription
epsdouble0.01gradient norm stopping criterion (stop when |g| < eps)
iterint100maximum number of iterations
optdictgd
Specific optimizer to be used.
bfgs:Broyden-Fletcher-Goldfarb-Shann
bfgs2:Broyden-Fletcher-Goldfarb-Shann (most efficient version)
cg-fr:Flecher-Reeves conjugate gradient algorithm
gd:Gradient descent.
simplex:Simplex algorithm of Nelder and Mead
cg-pr:Polak-Ribiere conjugate gradient algorithm
stepdouble0.001initial step size
toldouble0.1some tolerance parameter

nlopt

Minimizer algorithms using the NLOPT library, for a description of the optimizers please see 'http://ab-initio.mit.edu/wiki/index.php/NLopt_Algorithms'. Supported parameters are:

NameTypeDefaultDescription
ftoladouble0Stopping criterion: the absolute change of the objective value is below this value
ftolrdouble0Stopping criterion: the relative change of the objective value is below this value
local-optdictnone
local minimization algorithm that may be required for the main minimization algorithm.
gn-orig-direct-l:Dividing Rectangles (original implementation, locally biased)
gn-direct-l-noscal:Dividing Rectangles (unscaled, locally biased)
gn-isres:Improved Stochastic Ranking Evolution Strategy
ld-tnewton:Truncated Newton
gn-direct-l-rand:Dividing Rectangles (locally biased, randomized)
ln-newuoa:Derivative-free Unconstrained Optimization by Iteratively Constructed Quadratic Approximation
gn-direct-l-rand-noscale:Dividing Rectangles (unscaled, locally biased, randomized)
gn-orig-direct:Dividing Rectangles (original implementation)
ld-tnewton-precond:Preconditioned Truncated Newton
ld-tnewton-restart:Truncated Newton with steepest-descent restarting
gn-direct:Dividing Rectangles
ln-neldermead:Nelder-Mead simplex algorithm
ln-cobyla:Constrained Optimization BY Linear Approximation
gn-crs2-lm:Controlled Random Search with Local Mutation
ld-var2:Shifted Limited-Memory Variable-Metric, Rank 2
ld-var1:Shifted Limited-Memory Variable-Metric, Rank 1
ld-mma:Method of Moving Asymptotes
ld-lbfgs-nocedal: 
ld-lbfgs:Low-storage BFGS
gn-direct-l:Dividing Rectangles (locally biased)
none:don't specify algorithm
ln-bobyqa:Derivative-free Bound-constrained Optimization
ln-sbplx:Subplex variant of Nelder-Mead
ln-newuoa-bound:Derivative-free Bound-constrained Optimization by Iteratively Constructed Quadratic Approximation
ln-praxis:Gradient-free Local Optimization via the Principal-Axis Method
gn-direct-noscal:Dividing Rectangles (unscaled)
ld-tnewton-precond-restart:Preconditioned Truncated Newton with steepest-descent restarting
maxiterint100Stopping criterion: the maximum number of iterations
optdictld-lbfgs
main minimization algorithm
gn-orig-direct-l:Dividing Rectangles (original implementation, locally biased)
g-mlsl-lds:Multi-Level Single-Linkage (low-discrepancy-sequence, require local gradient based optimization and bounds)
gn-direct-l-noscal:Dividing Rectangles (unscaled, locally biased)
gn-isres:Improved Stochastic Ranking Evolution Strategy
ld-tnewton:Truncated Newton
gn-direct-l-rand:Dividing Rectangles (locally biased, randomized)
ln-newuoa:Derivative-free Unconstrained Optimization by Iteratively Constructed Quadratic Approximation
gn-direct-l-rand-noscale:Dividing Rectangles (unscaled, locally biased, randomized)
gn-orig-direct:Dividing Rectangles (original implementation)
ld-tnewton-precond:Preconditioned Truncated Newton
ld-tnewton-restart:Truncated Newton with steepest-descent restarting
gn-direct:Dividing Rectangles
auglag-eq:Augmented Lagrangian algorithm with equality constraints only
ln-neldermead:Nelder-Mead simplex algorithm
ln-cobyla:Constrained Optimization BY Linear Approximation
gn-crs2-lm:Controlled Random Search with Local Mutation
ld-var2:Shifted Limited-Memory Variable-Metric, Rank 2
ld-var1:Shifted Limited-Memory Variable-Metric, Rank 1
ld-mma:Method of Moving Asymptotes
ld-lbfgs-nocedal: 
g-mlsl:Multi-Level Single-Linkage (require local optimization and bounds)
ld-lbfgs:Low-storage BFGS
gn-direct-l:Dividing Rectangles (locally biased)
ln-bobyqa:Derivative-free Bound-constrained Optimization
ln-sbplx:Subplex variant of Nelder-Mead
ln-newuoa-bound:Derivative-free Bound-constrained Optimization by Iteratively Constructed Quadratic Approximation
auglag:Augmented Lagrangian algorithm
ln-praxis:Gradient-free Local Optimization via the Principal-Axis Method
gn-direct-noscal:Dividing Rectangles (unscaled)
ld-tnewton-precond-restart:Preconditioned Truncated Newton with steepest-descent restarting
ld-slsqp:Sequential Least-Squares Quadratic Programming
stepdouble0Initial step size for gradient free methods
stopdouble-infStopping criterion: function value falls below this value
xtoladouble0Stopping criterion: the absolute change of all x-values is below this value
xtolrdouble0Stopping criterion: the relative change of all x-values is below this value

Plugin consumers:

mia-2dgroundtruthreg mia-2dimageregistration mia-2dmyoica-nonrigid mia-2dmyoica-nonrigid2 mia-2dmyoica-nonrigid-parallel mia-2dmyomilles mia-2dmyoperiodic-nonrigid mia-2dmyopgt-nonrigid mia-2dmyoserial-nonrigid mia-3dmany2one-nonrigid mia-3dmotioncompica-nonrigid mia-3dnonrigidreg mia-3dnonrigidreg-alt mia-3dprealign-nonrigid mia-3drigidreg mia-3dserial-nonrigid

Prev Up 
Plugin type: mesh/io Home