LICORS (version 0.2.0)

mixed_LICORS: Mixed LICORS: An EM-like Algorithm for Predictive State Space Estimation

Description

mixed_LICORS is the core function of this package as it estimates the ``parameters'' in the model for the spatio-temporal process. $$ P(X_1, \ldots, X_{\tilde{N}}) \propto \prod_{i=1}^{N} P(X_i \mid \ell^{-}_i) = \prod_{i=1}^{N} P(X_i \mid \epsilon(\ell^{-}_i)) . $$

Usage

mixed_LICORS(LCs = list(PLC = NULL, FLC = NULL, dim = list(original = NULL, 
    truncated = NULL)), num.states.init = NULL, initialization = NULL, 
    control = list(max.iter = 500, alpha = 0.01, trace = 0, lambda = 0, 
        sparsity = "stochastic", CV.split.random = FALSE, CV.train.ratio = 0.75, 
        seed = NULL, loss = function(x, xhat) mean((x - xhat)^2), 
        estimation.method = list(PLC = "normal", FLC = "nonparametric")))

Arguments

LCs

list of PLCs and FLCs matrices (see output of data2LCs for details and formatting).

num.states.init

number of states to start the EM algorithm

initialization

a a) character string, b) vector, or c) matrix. a) results num.states.init many states initialized by passing the character string as method argument of initialize_states; if b) the vector will be taken as initial state labels; if c) the matrix will be taken as initial weights. Note that for both b) and c) num.states.init will be ignored. \(k = 1, \ldots, K\) of PLC \(i\)

control

a list of control settings for the EM algorithm. See complete_LICORS_control for details.

Value

An object of class "LICORS".

See Also

plot.mixed_LICORS, summary.mixed_LICORS

Examples

# NOT RUN {
data(contCA00)

LC_geom <- setup_LC_geometry(speed = 1, horizon = list(PLC = 2, FLC = 0), 
    shape = "cone")
bb <- data2LCs(t(contCA00$observed), LC.coordinates = LC_geom$coordinates)

mm <- mixed_LICORS(bb, num.states.init = 15, init = "KmeansPLC", 
    control = list(max.iter = 50, lambda = 0.001))
plot(mm)
ff_new <- estimate_LC_pdfs(bb$FLC, weight.matrix = mm$conditional_state_probs, 
    method = "nonparametric")
matplot(bb$FLC, ff_new, pch = ".", cex = 2)
# }