R implementation of the algorithm proposed by Mostafavi et al. (2012). This is a port of the MATLAB code provided by the authors, available here. For testing purposes, the example dataset contained in the link has been incorporated into threepRop as a dataset called example_dta.
Additionally, the package provides a function to efficiently simulate Stochastic Block Models for two classes, whose output can be used to test 3prop (view examples in the help page of the main function three_prop_cv). Also, threepRop contains a utiliy for calculating the area under the ROC curve (AUROC) for performance assessment.
For completeness, we also provide a simple implementation of the Generic Label Propagation (GLP) algorithm in function glp, as described by Equation 1 in Mostafavi et al. (2012).
The implementation relies on sparse data classes of package Matrix for the linear algebra operations needed on sparse matrices and vectors. Both three_prop_cv and glp only accept inputs from these classes, and will throw errors when used with dense objects.
if (!require(devtools)) {
install.packages('devtools')
}
devtools::install_github('miguelbiron/threepRop')Mostafavi, S., Goldenberg, A., & Morris, Q. (2012). Labeling nodes using three degrees of propagation. PloS one, 7(12), e51947.