README.md

In dchen49/GA: Variable Selection via Genetic Algorithm

Stat 243 Final Project

Variable Selection via Genetic Algorithm

Work in Progress

• David Chen
• Qi Chen
• Emily Suter
• Xinyi(Cindy) Zhang

Select() Outline

• Select (y, x, tbd, ...)
• initialize(p = initial population size, n = ncol(x))
  • make a p x n matrix of random Bernoullis
    • -> population = p x n matrix
• While (count < max iterations)
  • regress (x, y, model, fitness function, ...) # apply(population, regress)
  • Performs regression
  • Calculate fitness
    • -> Output: vector of fitness metrics
  • stop (regressOutput, fitness)
  • Test that the current generation fitness improves upon the last by some threshold
    • -> Output: Boolean (TRUE if improvement insignificant)
  • if (Stop==TRUE) Break
  • mate (population, fitness, survivor percentage, selection method/params, population size)
  • selects high fitness genotypes via selection method
    • -> Output: elite population
  • evolve(population, mutation prob, crossing prob, # of cross locations)
  • crossing(population, crossing prob, # of cross locations)
    • assign pairs
    • decide whether or not to cross, and crossing locations
      • -> Output: populationCrossed
  • mutate(population, mutation prob)
    • picks locations and changes values
    • -> Output: populationCrossedMutated
  • counter +1
  • pass new population back into while loop
• Output: top genotypes and fitness values!

dchen49/GA documentation built on May 3, 2019, 6:43 p.m.