Cpp function: fitting NEMoE paramters. (observed likelihood, penalized likelihood, complete likelihood, penalized complete likelihood) in all levels.

Cpp function: fitting NEMoE paramters. (observed likelihood, penalized likelihood, complete likelihood, penalized complete likelihood) in all levels.

fitNEMoE0(
  X,
  seg,
  Z,
  y,
  K,
  lambda1,
  lambda2,
  alpha1,
  alpha2,
  V_init,
  W_init,
  beta_max,
  EM_opt,
  itmax,
  itmin,
  adapt,
  btr,
  stop_all,
  verbose,
  early_stop = FALSE
)

Arguments

X	an aggregated data matrix (n*P) of input in experts network (rbind of all levels input).
seg	an integer vector of length in each level (L).
Z	a data matrix (n*q) of input in gating network.
y	a vector of response (n).
K	A number of latent class.
lambda1	aggregated penalty lambda parameters in experts network(P*K).
lambda2	penalty parameters lambda in gating network(q*1).
alpha1	penalty parameters alpha in experts network(L*K).
alpha2	penalty parameter alpha in gating network(1).
V_init	initial parameters in gating network((q+1)*K).
W_init	aggregated initial parameters in experts network (cbind of all levels) ((P+L)*K).
beta_max	A number of maximal of coefficients to avoid divergence of during the fitting.
EM_opt	A integer indicate methods for EM algorithm. 0="EM", 1 = "CEM", 2 = "SEM", 3= "SAEM", 4="GEM".
itmax	maximal numbers of iteration in fitting NEMoE.
itmin	minimal numbers of iteration in fitting NEMoE.
adapt	A boolean variable indicates whether to use adaptive NEMoE.
btr	A boolean variable indicates whether to use backtracking during fitting NEMoE.
stop_all	A boolean variable indicates whether to stop by (likelihood converge)&(parameters converge)
verbose	A boolean variable indicates whether to show PLL during each iteration.
early_stop	A boolean variable indicates whether to stop when one latent class have select zero variables (to save time).

Value

A matrix of fitting result.