scalefreeForest: Scale-free forest density estimation

In zhejosephliu/scalefreeForest: Scale-free forest density estimation

Description

The main function for scale-free forest density estimation.

Usage

scalefreeForest(xtrain, xheld, lambda = seq(0.005, 0.15, 0.005), iter.max = 100, range = NULL, verbose = TRUE)

Arguments

xtrain	An n by d data matrix as training set, where n is the sample size and d is the dimension.
xheld	An m by d data matrix as held-out set, where m is the sample size and d is the dimension.
lambda	A sequence of positive numbers to control the regularization of the log degree penalty. The default value is seq(0.005, 0.15, 0.005).
iter.max	The maximal number of steps in the iterative reweighted Kruskal's algorithm. The default value is 100.
range	The range for each of the variables. The default value is NULL.
verbose	If verbose = FALSE, tracing information printing is disabled. The default value is TRUE.

Details

The training data set is used to construct scale-free forests and the corresponding density estimators, while the held-out data set is then used to determine the optimal scale-free forest by maximizing the held-out log-likelihood.

Value

loglike	Maximal held-out log-likelihood for the optimal scale-free forest density estimators corresponding to the regularization parameters.
adj	Adjacency matrices of the optimal scale-free forests corresponding to the regularization parameters.
best.loglike	Maximal held-out log-likelihood over all the regularization parameters.
best.adj	Adjacency matrix of the optimal scale-free forest corresponding to the maximum of the held-out log-likelihood over all the regularization parameters.

Examples

library(igraph)

fit <- scalefreeForest(xtrain, xheld)
sf <- fit$best.adj
plot(sf)

zhejosephliu/scalefreeForest documentation built on May 4, 2019, 10:17 p.m.