coupled4_kernel: Runs a 4-coupled Markov kernel

In jeremyhengjm/UnbiasedGradients: Unbiased estimators for partially observed diffusions

Runs a 4-coupled Markov kernel

Description

Runs four coupled kernels (two at each discretization level) that leaves smoothing distribution at corresponding level invariant.

Usage

coupled4_kernel(
  model,
  theta,
  discretization,
  observations,
  nparticles,
  resampling_threshold,
  coupled_resampling,
  ref_trajectory_coarse1,
  ref_trajectory_coarse2,
  ref_trajectory_fine1,
  ref_trajectory_fine2,
  algorithm = "CPF",
  treestorage = FALSE
)

Arguments

model	a list representing a hidden Markov model, e.g. hmm_ornstein_uhlenbeck
theta	a vector of parameters as input to model functions
discretization	lists containing stepsize, nsteps, statelength, obstimes for fine and coarse levels, and coarsetimes of length statelength_fine indexing time steps of coarse level
observations	a matrix of observations, of size nobservations x ydimension
nparticles	number of particles
resampling_threshold	ESS proportion below which resampling is triggered (always resample at observation times by default)
coupled_resampling	a 4-marginal coupled resampling scheme, such as coupled4_maximalchains_maximallevels_independent_residuals
ref_trajectory_coarse1	a matrix of first reference trajectory for coarser discretization level, of size xdimension x statelength_coarse
ref_trajectory_coarse2	a matrix of second reference trajectory for coarser discretization level, of size xdimension x statelength_coarse
ref_trajectory_fine1	a matrix of first reference trajectory for finer discretization level, of size xdimension x statelength_fine
ref_trajectory_fine2	a matrix of second reference trajectory for finer discretization level, of size xdimension x statelength_fine
algorithm	character specifying type of algorithm desired, i.e. CPF for conditional particle filter, CASPF for conditional ancestor sampling particle filter, CBSPF for conditional backward sampling particle filter
treestorage	logical specifying tree storage of Jacob, Murray and Rubenthaler (2013); if missing, this function store all states and ancestors

Value

four new trajectories stored as matrices of size xdimension x statelength_coarse/fine.

jeremyhengjm/UnbiasedGradients documentation built on Nov. 19, 2023, 11:24 p.m.