.internal/design.md

In erickim/GA: Genetic Algorithms for Regression

Stat 243 Final Project: Genetic Algorithms Design Doc

select

The main function is select(...) located in the select.R file.

It will take arguments: data - the data, should be of class data.frame. reg_type - the type of regression, a string of 'lm', 'glm', and perhaps 'glmnet'. family - the family to be passed into 'glm' or perhaps 'glmnet'. objective - the objective function default AIC but can also allow for user implemented functions. fitness - similar to objective but can be different, see Givens/Hoeting for details. selection - the selection mechanism, probably a string, see Givens/Hoeting for details. crossover - the crossover type. mutation - the mutation rate. P - the population size. seed - the seed to start the process, defaulted at 1.

initialize

There should be a function to initialize the algorithm.

It will take arguments as follows that were passed into select Y - the response vector. X - the feature matrix * P - the population size.

crossover

There should be a function to implement crossover.

It will take arguments: parent1 and parent2 - two candidate solutions (a vector of 0's and 1's). type - see page 77 of Givens/Hoeting (either single or multiple) * num_splits - if type was multiple, then the user supplied number of splits. Will default to 2 if invalid num_splits passed.

mutate

There should be a function ot implement mutation after crossover.

It will take arguments: offspring - an output from crossover. rate - the mutation rate passed into select.

selection

There should be a function to implement selection and update the candidate solutions at each stage. It will call crossover and mutation.

It will take arguments: type - either 'oneprop' or 'twoprop' as described in Page 76 of Givens/Hoeting pop_fitness - the candidate fitness scores

erickim/GA documentation built on May 28, 2019, 7:15 a.m.