mutual_info: Mutual Information
In baruuum/btoolbox: random toolbox

Description Usage Arguments Details References Examples

Calculates Variants of the mutual information of the rows and the columns of a cross table

1	mutual_info(x, normalized = "none", adjusted = FALSE)

`x`	a cross table of frequencies; has to be a integer matrix or table object
`normalized`	a string that specifies how the mutual information should be normalized. Must be one of either "none", "min", "max", "arithmetic", "geometric", or "harmonic."
`adjusted`	if TRUE, calculates the adjusted version based on the hypergeometric model of randomness. Defaults to FALSE

The normalized option specifies how the row and column entropies are used to normalize the mutual information. For example, with "min", the normalized mutual information (NMI) would be I(R, C) / \min[H(R), H(C)], where I(R, C) is the mutual information between the row and the column, and H(R) and H(C), respectively, are entropies of the row and column marginals. When adjusted is set to TRUE, the adjusted mutual information, as suggested by Vinh et al. (2010), is returned. If what == "none" and adjusted == TRUE, the function returns I(R,C) - E[I(R,C)].

Nguyen Xuan Vinh, Julien Epps, and James Bailey. 2010. "Information Theoretic Measures for Clusterings Comparison: Variants, Properties, Normalization and Correction for Chance," Journal of Machine Learning Research 11: 2837-2854.

data(sampson3, package = "btoolbox")

info_part = igraph::cluster_infomap(sampson3)
groups = igraph::vertex_attr(sampson3, "group")
print(tab <- table(groups, info_part$membership))

mutual_info(tab, normalize = "geometric", adjusted = TRUE)