alineR (version 1.1.4)

optimize.features: Supervised learning with a genetic algorithm

Description

Runs a genetic algorithm to find optimal parameter settings based on expert alignment determinations.

Usage

optimize.features(set, ranking, num = 200, step = 45, replication = 5, 
    list = FALSE)

Arguments

set

the output from set.generation function, which is a two element list containing the original word pairs and possible alignments.

ranking

a vector specifying the correct alignment in the candidate alignments generated.

num

number of populations in the genetic algorithm

step

number of iterations in the genetic algorithm

replication

number of independent genetic algorithm optimizations.

list

Whether or not to return the entire result of the genetic algorithm which contains a big list of possible parameters and corresponding performance in each independent replication

Value

If list=FALSE, the function returns a single vector representing the optimal parameter values.

If list=TRUE, the function returns a list where each top-level element corresponds to the number of replications. Within each replicate, two elements are returned:

performance

Performance values for each population.

optimized_parameters

Feature values at each step in the optimization process.

Examples

# NOT RUN {
# This simplified example illustrates the supervised learning workflow 
# some cognate data
data<-data.frame(dog=c('dog','perro'),cat=c('cat','gato'),rat=c('rat','rata'))

# generate training data for linguist (not written)
M1<-generate.training(raw.data=data, search.size=100)

# optimize features using expert determinations: 1,1,1
optimize.features(set=M1, ranking=c(1,1,1),
            num=20, step=5, replication=2, list=FALSE)

# }