Analysing and comparing problem landscapes for black-box optimization via length scale
Rachael Morgan · The University of Queensland · 2015
Optimization problems are of fundamental practical importance and can be found in almost every aspect of human endeavour. Yet remarkably, we have a very limited understanding of the nature of optimization problems and subsequently of how, why and when different algorithms perform well or poorly. The notion of a problem landscape captures the relation-ship between the objective function and the problem variables. It is clear that the structure of this landscape is vital in understanding optimization problems, however analysis of this structure presents major challenges. Apart from the often high-dimensionality of problem landscapes, information available in the black-box setting is limited to the solutions in the feasible search space and their respective objective function values. Landscape analysis has received some attention in the optimization literature, mainly in evolutionary computation. However there are some important limitations of this work as well as many open issues around its practical utility. This thesis proposes a novel framework and practical techniques for the analysis of op-timization problems utilizing information available in the black-box setting. The concept of