GABi (version 0.1)

getFitnesses.entropy: Entropy-based Bicluster Fitness Function

Description

A fitness function for the GABi biclustering framework, based on the principle that the less likely a bicluster would be observed by chance, the more interesting it is to discover that bicluster.

Usage

getFitnesses.entropy(chr)

Arguments

chr

Numeric vector representing a GA solution to the biclustering problem (i.e. a subset of the columns from x across which to look for the pattern).

Value

chr.

Details

A fitness function is fundamental to the success of a GA. In this case, getFitnesses.entropy evaluates the desirability of biclusters by estimating the probability of a given selection of the columns from dataset x (argument for function GABi) displaying a consistent block of 1s involving the features that are observed to fit this pattern. Makes use of fitnessArgs a list of parameters in the environment of execution of the biclustering function GABi. Notably, the element featureWeights is a numeric vector encoding the probability of any randomly selected column of the input matrix x having a high value of the corresponding row. This is used in the entropy calculation for the corresponding bicluster. And the element consistency is used to apply a stringency threshold for selecting features (i.e. only those with the proportion of high values across the subset of samples being greater than consistency).