ofmlr-package: Fit online finite mixtures of logistic regression models

In Nth-iteration-labs/ofmlr: Fit online finite mixtures of logistic regression models

Description

The ofmlr package allows one to fit, online (or in a data stream) an multiple finite mixtures of logistic regression models. The main workhorse online_log_mixture and its utility functions (such as add_observation can be used to fit a model. The multi_online_log_mixture object allows one to fit multiple models in paralel and compare their outcomes.

ofmlr functions

Next to standards like plot and summary the package implements the folling functions

• online_log_mixture: Constructor for new ofmlr object.

• multi_online_log_mixture: Constructor for an object containing multiple ofmlr models.

• add_model: Add additional models to a multi_online_log_mixture object.

• add_observation: Add observations to a model and update parameters

The package also contains the following utility functions:

• generate_mixture: Utility function to create a mixture dataset.

• generate_probabilty_vector: Function to create a probablity vector.

• inv_logit: The inverse logit function.

Examples

M1 <- online_log_mixture(3,3, trace=1)
for(i in 1:10000){ 	
	X <- runif(3,-2,2)
	y <- rbinom(1, 1, inv_logit(c(0,-2,2)%*%X))
	M1 <- add_observation(M1, y, X, 0)
}
plot(M1, params=TRUE)

Nth-iteration-labs/ofmlr documentation built on Sept. 27, 2020, 9:49 p.m.