gpnlreg.estim.def: Locally estimates Matern and rigid transformation parameters...

In ashtonwiens/gp.nonrigid.registration: An R package for rigid and nonrigid registration of spatial surfaces

Description

gpnlreg.estim.def takes a gpnl.obj and a list of estimation and fitting parameters, and using either a serial or an Rmpi implementation, calls parallel.fcn on every window of data in gpnl.obj$data.subsets, returning a gpnl.obj with an est.results component including all of the results of the estimation and model fitting.

Usage

gpnlreg.estim.def(gpnl.obj, est.pars)

Arguments

gpnl.obj

a gpnl.obj containing a data.subsets component

est.pars

a list containing n.subsamples (the number of subsamples to be taken from each data set in each window), smoothness (of the Matern covariance function in the local estimation procedure), lambd (value of tuning parameter in the penalty in the objective function), penalty.type (one of 'L1', 'L2', 'elastic', or 'none'), upper.bounds and lower.bounds (box constraints in the local likelihood parameter estimation), min.num.samples (the minimum number of samples needed in both data sets to perform local estimation on a window of data), and est.type (a character vector 'serial' or 'Rmpi' indicating how the estimation should be performed).

Value

a gpnl.obj containing an est.results component

See Also

serial.implementation rmpi.implementation parallel.fcn

ashtonwiens/gp.nonrigid.registration documentation built on April 11, 2020, 12:28 a.m.