R-chaosoptimiser: an optimiser for unconstrained global nonlinear optimisation written in R language for statistical computing

Abstract

This paper discusses using R-chaosoptimiser, an R language package for nonlinear optimisation based on gradient techniques and chaos optimisation algorithms. Its implementation was based on three building blocks which could be executed alone or un combination: the first carrier wave algorithm, the chaos-based cyclical coordinate search method and the second wave carrier algorithm. Using chaos optimisation algorithms allows the tool to break away from local optimal points and converge towards an overall optimum inside a predefined search domain. Within the previous components, a user would be specifying the BFGS algorithm for refining the current best solution. Using the BFGS algorithm is not mandatory, so that its implementation was able to optimise problems having objective function discontinuities. However, the BFGS algorithm is a powerful local search method, meaning that it is used to exploit current knowledge about an objective function for improving a current solution; an explanatory example is presented.

Keywords

• optimisation
• R language
• gradient-based method
• chaos
• algorithm