RANKS (version 1.1)

label.prop: Label propagation

Description

Function that implements the Label propagation algorithm of Zhu and Ghahramani

Usage

label.prop(W, ind.positives, tmax = 1000, eps = 1e-05, norm = TRUE)

Value

A list with three elements:

p

numeric vector. Scores of each node at the steady state or after tmax iterations

ind.positives

indices of the "core" positive examples of the graph (it is equal to the same input parameter)

n.iter

number of performed steps/iterations

Arguments

W

a numeric matrix representing the adjacency matrix of the graph

ind.positives

indices of the "core" positive examples of the graph. They represent the indices of W corresponding to the positive examples

tmax

maximum number of iterations (def: 1000)

eps

numeric. Maximum allowed difference between the computed probabilities at the steady state (def. 1e-5)

norm

boolean. If TRUE (def) the adjacency matrix \(W\) of the graph is normalized to \(M = D^{-1} * W\), otherwise it is assumed that the matrix \(W\) is just normalized

Details

label.prop implements the label propagation algorithm on a given graph by performing 1 or more steps on the graph, depending on the value of the tmax parameter. It stops also if the difference of the norm of the scores between two consecutive steps is less than eps.

References

Zhu, X., Ghahramani, Z., Lafferty, J.: Semi-Supervised Learning Using Gaussian Fields and Harmonic Functions. In: Proc. of the Twentieth International Conference on Machine Learning, Washington DC (2003) 912-919

Examples

# Application of label prop algorithm to the prediction of the DrugBank category Penicillins
# using the Tanimoto chemical structure similarity network 
# between 1253 DrugBank drugs
library(bionetdata);
data(DD.chem.data);
data(DrugBank.Cat);
labels <- DrugBank.Cat[,"Penicillins"];
ind.pos <- which(labels==1);
label.prop(DD.chem.data, ind.pos, tmax = 10, eps = 1e-05, norm = TRUE);