fit.lsa_bcsgplvm: Fit a large scale approximation back constrained structured...

In mattdneal/GPLVM: Gaussian Process Latent Variable Models

Description

Fit a large scale approximation back constrained structured GPLVM model

Usage

fit.lsa_bcsgplvm(X, q = 2, iterations = 1000, plot.freq = 100,
  classes = 1, Z.init = NULL, A.init = NULL, K.bc.l = "auto",
  K.bc.l.selection.params = NULL, K.bc.l.plot.graphs = T,
  Z.prior = c("normal", "uniform", "discriminative"), par.init = NULL,
  points.in.approximation = 1024, optimization.method = c("SMD", "ADAM"),
  optimization.method.pars = NULL, parameter.opt.iterations = 300,
  par.fixed.par.opt = NULL, par.fixed.A.opt = NULL, verbose = FALSE,
  subsample.flat.X = NULL, Z.prior.params = list(), save.X = FALSE,
  optimize.structure.params.first = TRUE, optimize.all.params = FALSE,
  ivm = FALSE, ivm.selection.size = 2048)

Arguments

Z.init	Either a matrix of initial latent values, or "PCA" for PCA start values, or ISOMAP for ISOMAP start values.
K.bc.l	lengthscale to use for backconstraints. Either a numeric value or "auto", which automatically determines an appropriate lengthscale.
K.bc.l.selection.params	parameters for algorithm which automatically selects backconstraint lengthscale. See select.bc.l.centile for details.
K.bc.l.plot.graphs
par.init	Vector of parameters: alpha, sigma, l_Z, followed by the lengthscales for the structural dimensions
ivm	select points in each step using IVM
ivm.selection.size	number of points to consider for each IVM selection, NULL for all points

mattdneal/GPLVM documentation built on May 7, 2019, 1:26 p.m.