GenAlg: A generic Genetic Algorithm for feature selection

Description

These functions allow you to initialize (GenAlg) and iterate (newGeneration) a genetic algorithm to perform feature selection for binary class prediction in the context of gene expression microarrays or other high-throughput technologies.

Usage

GenAlg(data, fitfun, mutfun, context, pm=0.001, pc=0.5, gen=1)
newGeneration(ga)
popDiversity(ga)

Value

Both the GenAlg generator and the newGeneration functions return a GenAlg-class object. The popDiversity function returns a real number representing the average diversity of the population. Here diversity is defined by the number of alleles (selected features) that differ in two individuals.

Arguments

data: The initial population of potential solutions, in the form of a data matrix with one individual per row.
fitfun: A function to compute the fitness of an individual solution. Must take two input arguments: a vector of indices into rows of the population matrix, and a context list within which any other items required by the function can be resolved. Must return a real number; higher values indicate better fitness, with the maximum fitness occurring at the optimal solution to the underlying numerical problem.
mutfun: A function to mutate individual alleles in the population. Must take two arguments: the starting allele and a context list as in the fitness function.
context: A list of additional data required to perform mutation or to compute fitness. This list is passed along as the second argument when fitfun and mutfun are called.
pm: A real value between 0 and 1, representing the probability that an individual allele will be mutated.
pc: A real value between 0 and 1, representing the probability that crossover will occur during reproduction.
gen: An integer identifying the current generation.
ga: An object of class GenAlg

Author

Kevin R. Coombes krc@silicovore.com, P. Roebuck proebuck@mdanderson.org

Examples

Run this code

# generate some fake data
nFeatures <- 1000
nSamples <- 50
fakeData <- matrix(rnorm(nFeatures*nSamples), nrow=nFeatures, ncol=nSamples)
fakeGroups <- sample(c(0,1), nSamples, replace=TRUE)
myContext <- list(dataset=fakeData, gps=fakeGroups)

# initialize population
n.individuals <- 200
n.features <- 9
y <- matrix(0, n.individuals, n.features)
for (i in 1:n.individuals) {
  y[i,] <- sample(1:nrow(fakeData), n.features)
}

# set up the genetic algorithm
my.ga <- GenAlg(y, selectionFitness, selectionMutate, myContext, 0.001, 0.75)

# advance one generation
my.ga <- newGeneration(my.ga)