dynamic_importance_sampling: Runs belief update from a piece of dynamic evidence over a...

In ramidst: An Interface to the AMIDST Toolbox for Data Stream Processing

Description

Runs belief update from a piece of dynamic evidence over a dynamic Bayesian network

Usage

dynamic_importance_sampling(dyn_network, target_variable, dyn_evidence,
  sample_size = 50)

Arguments

dyn_network	a java object of class DynamicBayesianNetwork over which the computations will be carried out
target_variable	the name of the variable over which the posterior distribution will be computed
dyn_evidence	the observations over the dynamic Bayesian network.
sample_size	the size of the sample used to estimate the posterior distribution.

Value

a data.frame with the posterior distribution over the target variable on the different time slices.

Note

The function uses importance sampling and is based on the factored frontier method, see the AMIDST toolbox documentation.

Examples

network <- dbn_generator(1,2,2)
print_amidst_bn(network)
stream <- generate_stream_from_dbn(network,1,10,"ClassVar")
resultsIS <- dynamic_importance_sampling(network,"ClassVar",stream)
plot(resultsIS[,2],type="l",ylim = c(0,1),col="red",xlab="Time slice",ylab="Prob. ClassVar = 1")

ramidst documentation built on May 30, 2017, 5:27 a.m.