1% Generated by roxygen2: do not edit by hand 2% Please edit documentation in R/global.R 3\name{isres} 4\alias{isres} 5\title{Improved Stochastic Ranking Evolution Strategy} 6\usage{ 7isres(x0, fn, lower, upper, hin = NULL, heq = NULL, maxeval = 10000, 8 pop.size = 20 * (length(x0) + 1), xtol_rel = 1e-06, 9 nl.info = FALSE, ...) 10} 11\arguments{ 12\item{x0}{initial point for searching the optimum.} 13 14\item{fn}{objective function that is to be minimized.} 15 16\item{lower, upper}{lower and upper bound constraints.} 17 18\item{hin}{function defining the inequality constraints, that is 19\code{hin>=0} for all components.} 20 21\item{heq}{function defining the equality constraints, that is \code{heq==0} 22for all components.} 23 24\item{maxeval}{maximum number of function evaluations.} 25 26\item{pop.size}{population size.} 27 28\item{xtol_rel}{stopping criterion for relative change reached.} 29 30\item{nl.info}{logical; shall the original NLopt info been shown.} 31 32\item{...}{additional arguments passed to the function.} 33} 34\value{ 35List with components: 36 \item{par}{the optimal solution found so far.} 37 \item{value}{the function value corresponding to \code{par}.} 38 \item{iter}{number of (outer) iterations, see \code{maxeval}.} 39 \item{convergence}{integer code indicating successful completion (> 0) 40 or a possible error number (< 0).} 41 \item{message}{character string produced by NLopt and giving additional 42 information.} 43} 44\description{ 45The Improved Stochastic Ranking Evolution Strategy (ISRES) algorithm for 46nonlinearly constrained global optimization (or at least semi-global: 47although it has heuristics to escape local optima. 48} 49\details{ 50The evolution strategy is based on a combination of a mutation rule (with a 51log-normal step-size update and exponential smoothing) and differential 52variation (a Nelder-Mead-like update rule). The fitness ranking is simply 53via the objective function for problems without nonlinear constraints, but 54when nonlinear constraints are included the stochastic ranking proposed by 55Runarsson and Yao is employed. 56 57This method supports arbitrary nonlinear inequality and equality constraints 58in addition to the bound constraints. 59} 60\note{ 61The initial population size for CRS defaults to \code{20x(n+1)} in 62\code{n} dimensions, but this can be changed; the initial population must be 63at least \code{n+1}. 64} 65\examples{ 66 67### Rosenbrock Banana objective function 68fn <- function(x) 69 return( 100 * (x[2] - x[1] * x[1])^2 + (1 - x[1])^2 ) 70 71x0 <- c( -1.2, 1 ) 72lb <- c( -3, -3 ) 73ub <- c( 3, 3 ) 74 75isres(x0 = x0, fn = fn, lower = lb, upper = ub) 76 77} 78\references{ 79Thomas Philip Runarsson and Xin Yao, ``Search biases in 80constrained evolutionary optimization,'' IEEE Trans. on Systems, Man, and 81Cybernetics Part C: Applications and Reviews, vol. 35 (no. 2), pp. 233-243 82(2005). 83} 84\author{ 85Hans W. Borchers 86} 87