select: select()

In jnieser/GeneticAlgorithmRPackage: Genetic Algorithm

select()

Description

Find the best covariates in X such that the regression on y by a specific type has a low fitness value.

Usage

select(
  X,
  y,
  fit.function = AIC,
  glm.family = "gaussian",
  operator = NULL,
  pop.size = 100,
  select.type = "tournament",
  mutate.rate = NULL,
  min_iter = 5,
  max_iter = 100,
  converge.thres = 0.01,
  verbose = 0
)

Arguments

X	m by n matrix of X as covariate
y	list of numbers; the variable to be regressed on
fit.function	a user-defined function; this function will give the fitness value of any given gene and X, and y variable. By defaut, it is the AIC.
glm.family	String; user-defined regression type, e.g. Gaussian for lm, binomial for logistic
operator	a user-defined function; this function determines how to mutate the genes. By default it is NULL.
pop.size	integer; population size
select.type	String; type for selection, rank or tournament
mutate.rate	double; rate for mutation, default to be 0.01
min_iter	integer; minimum number of iterations
max_iter	integer; maximum number of iterations
converge.thres	double; threshold to determine whether to stop iterating early. If it is 0.01, that means if the ratio of the change of mean of fitness value from last generation and the mean of fitness value from last generation < 0.01, then we finish our iterations early, ignoring the max_iter.
verbose	integer; determine the type of log to be printed.

Author(s)

Jiayang Nie

Examples

library(GA)
X = matrix(rnorm(120*100, 10, 5), 120, 100)
y = rowSums(X)
X_noise = matrix(rnorm(120*100, -10, 1), 120, 100)
X = cbind(X, X_noise)
test1 <- select(X,y, max_iter = 100, verbose=2)
test2 <- select(X,y, max_iter = 100, select.type="rank", verbose=2)
test1$HallOfFame$genes[,1]

jnieser/GeneticAlgorithmRPackage documentation built on March 19, 2022, 5:35 a.m.