R/initval.R

In adana: Adaptive Nature-Inspired Algorithms for Hybrid Genetic Optimization

# Value encoded initialization 
initval = function(n, m, prevpop, lb, ub, nmode="real", type, ...){
  if(missing(m)){
    if(!missing(lb))
      m = length(lb)
    else
      m = 8 # Default gene count
  }
  if(missing(prevpop)) prevpop = NULL
  if(is.null(prevpop)){
    nprev = 0
  }else{
    prevpop = as.matrix(unname(prevpop))
    nprev = nrow(prevpop)
    m = ncol(prevpop)
  }
  if(missing(n))
    if(is.null(prevpop)) n = 4*m else n=nprev
    if(missing(lb)) lb = rep(0, m)
    if(missing(ub)) ub = rep(1, m)
    if(length(lb) != length(ub)) 
      stop("ub length must be equal to lb length!")
    if(missing(type)) type = 1
    initpop = matrix(NA, ncol=m+2, nrow=n) # Initial population
    if(nprev==0){
      for(i in 1:n) 
        for(j in 1:m)
          initpop[i,j] = runif(1, lb[j], ub[j]) 
    }else if(nprev<n){
      initpop[1:nprev,1:m] = prevpop[1:nprev,]
      for(i in (nprev+1): n) 
        for(j in 1:m)
          initpop[i,j] = runif(1, lb[j], ub[j]) 
    }else if(nprev>n){
      initpop[1:n,1:m] = prevpop[1:n,]
    }else{
      initpop[1:nprev,1:m] = prevpop[1:nprev,]
    }
    if(nmode=="integer") 
      initpop[,1:m] = apply(initpop[,1:m], 2, round)
    initpop[,m+1] = 0
    if(nprev==0){
      rnames = paste0("T0.", 1:n)
    }else if(nprev<n){
      rnames = c(paste0("Pre.", 1:nprev), 
                 paste0("T0.", 1:(n-nprev)))
    }else{
      rnames = paste0("Pre.", 1:n)
    }
    rownames(initpop) = rnames
    colnames(initpop) = c(paste0("gen", 1:m), "t", "fitval")
    if(type==2)
      initpop = initpop[,1:m]
    return(population=initpop)
}