StocAp: Estimation of potentials values via Stochastic Approximation.

In VicFreguglia/GibbsRF: Tools for Gibbs Random Fields models

Description

Estimates potentials vias Stochastic Approximation algorithm.

Usage

StocAp(X, gModel, type, initial = "MPLE", MC_size = 1, iter = 10000,
  macrosteps = 40, print_sample = TRUE, a0 = 1, a1 = 1/20,
  a2 = 3)

Arguments

X	The observed random field (matrix).
gModel	A GibbsModel object (V and vMat can be NULL).
type	Model type. Either "general", "symetric" or "equal". Check GibbsMPLE for details.
initial	Wich method to get initial estimates. Currently only "MPLE" available.
MC_size	Number of fields to sample at each step.
iter	Maximum number of iterations.
macrosteps	Number of macrosteps per random field simulation.
print_sample	Indicates if the a sample of the current model should be printed each 10 steps.
a0, a1, a2	constants to define size of each step. The gradient vector is multiplied by a0/(i*a1 + a2)

Value

A GibbsModel object estimated potentials.

Author(s)

Victor Freguglia Souza

Examples

StocAp(example.X,example.GibbsModel,"symetric",MC_size = 5)

VicFreguglia/GibbsRF documentation built on Oct. 25, 2019, 11:19 p.m.