VI_PMM_algo: Variational Inference Poisson Mixture

In advaitr8/PoissonMixR:

Description

This function allows you to run a VI algorithm for learning Poisson mixtures

Usage

VI_PMM_algo(x, K, iter = 1000, alpha = 0.1, a = 4.5, b = 0.25)

Arguments

x	this is the data
K	this is the number
iter	number of iterations defaults to 1000
alpha	hyperparameter of probability distribution of Poisson models defaults to 0.1
a	Gamma prior parameter defaults to 0.1
b	Gamma prior parameter defaults to 0.25

Examples

#generate some Poisson mixture data
data <- c(rpois(100,2), rpois(100,10))

#run the VI algorithm to learn the mixture model
t2 <- VI_PMM_algo(data,2,iter = 1000)

#plot the VI objective function, should be monotonically increasing and monitors convergence
plot(t2$L_vi,
     type = 'l',
     main = "convergence of VI objective function",
     bty = 'n')
     
#check the cluster assignments (using original data this is not a prediction, which can also be done)
cluster_indicator <- c()
for(i in 1:length(data)){
  cluster_indicator <- c(cluster_indicator,which.max(t2$norm_phi[i,]))
}
plot(data,cluster_indicator,
     pch = 16,
     bty = 'n')

advaitr8/PoissonMixR documentation built on May 28, 2019, 3:59 p.m.