mRMRe: mim – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

mim

Accessor function for the 'mim' information in a mRMRe.Data, mRMRe.Filter and mRMRe.Network object

Description

In both mRMRe.Filter and mRMRe.Network objects, a sparse mutual information matrix is computed for the mRMRe procedure and this lazy-evaluated matrix is returned. In the context of a a mRMRe.Data 'mim', the full pairwise mutual information matrix is computed and returned.

Usage

## S4 method for signature 'mRMRe.Data'
mim(object, prior_weight, continuous_estimator, outX, bootstrap_count)
## S4 method for signature 'mRMRe.Filter'
mim(object, method)
## S4 method for signature 'mRMRe.Network'
mim(object)

Arguments

`object`	a `mRMRe.Data, mRMRe.Filter or mRMRe.Network` object.
`prior_weight`	a numeric value [0,1] of indicating the impact of priors (mRMRe.Data only).
`continuous_estimator`	an estimator of the mutual information between features: either "pearson", "spearman", "kendall", "frequency" (mRMRe.Data only).
`outX`	a boolean used in the concordance index estimator to keep or throw out ties (mRMRe.Data only).
`bootstrap_count`	an integer indicating the number of bootstrap resampling used in estimation (mRMRe.Data only).
`method`	either "mi" or "cor"; the latter will return the correlation coefficients (rho) while the former will return the mutual information (-0.5 * log(1 - (rho^2))).

Author(s)

Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains

Examples

set.thread.count(2)
data(cgps)
feature_data <- mRMR.data(data =  data.frame(cgps.ge))

# Calculate the pairwise mutual information matrix
mim(feature_data)
filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5,
						feature_count = 2)

# Obtain the sparse (lazy-evaluated) mutual information matrix.
mim(filter)

mRMRe

"Parallelized Minimum Redundancy, Maximum Relevance (mRMR) Ensemble Feature Selection"

v2.1.0

Artistic-2.0

Authors

Nicolas De Jay, Simon Papillon-Cavanagh, Catharina Olsen, Gianluca Bontempi, Benjamin Haibe-Kains

Initial release

2020-01-08

mim