Introduction

This article illustrates how existing algorithms can be integrated into the SPOT framework. The SPOT package can be downloaded from the comprehensive R archive network at http://CRAN.R-project.org/package=SPOT. SPOT is one possible implementation of the sequential parameter optimization (SPO) framework introduced in Bartz-Beielstein (2006). For a detailed documentation of the functions from the SPOT package, the reader is referred to the package help manuals.

The performance of modern search heuristics such as evolution strategies(ES), differential evolution (DE), or simulated annealing (SANN) relies crucially on their parametrizations--or, statistically speaking, on their factor settings. The term algorithm design summarizes factors that influence the behavior (performance) of an algorithm, whereas problem design refers to factors from the optimization (simulation) problem. Population size in ES is one typical factor which belongs to the algorithm design, the search space dimension belongs to the problem design.

The paper is structured as follows: Section 2 presents an example how algorithms written in JAVA can be integrated into the SPOT framework. Two JAVA implementations of evolution strategies are discussed.

• A simple (1+1)-ES in Sec. 2.1
• Hansen's CMA-ES in Sec. 2.2