Generic Hardness Estimation using Fitness and Parameter Landscapes applied to Robust Taboo Search and the Quadratic Assignment Problem

Outline

E. Pitzer, A. Beham, M. Affenzeller - Generic Hardness Estimation using Fitness and Parameter Landscapes applied to Robust Taboo Search and the Quadratic Assignment Problem - Companion Publication of the 2012 Genetic and Evolutionary Computation Conference, GECCO'12 Companion, Philadelphia, Vereinigte Staaten von Amerika, 2012, pp. 393-400

Abstract

Fitness landscape analysis methods have become an increasingly popular topic for research. The future application of these methods to metaheuristics can yield advanced self-adaptive metaheuristics and knowledge bases that can take the role of expert systems in the field of optimization. One important feature of such an expert system would be the pre- diction of algorithm effort on a certain instance. Estimating whether a certain algorithm is able to tackle the problem adequately or not is a valuable piece of information that currently only an experienced human expert can give. The ability to generate such an advice automatically is, therefore, an important milestone. While fitness landscape analysis methods have been developed for exactly this purpose, it has been shown in the past that single-value analyses have limited applicability. Here, a general method for extracting fitness landscape features will be shown in combination with regression models that indicate a strong correlation between the actual and the predicted effort. Significant potential to increase the prediction quality arises when combining several measures each derived from several different sampling trajectories.