Barrier Trees on Poset-Valued Landscapes

  • Authors:
  • Peter F. Stadler;Christoph Flamm

  • Affiliations:
  • Institut fü/r Theoretische Chemie und Molekulare Strukturbiologie, Universitä/t Wien, Wä/hringerstrasse 17, A-1090 Wien, Austria/ Bioinformatik, Institut fü/r Informatik, Universit ...;Institut fü/r Theoretische Chemie und Molekulare Strukturbiologie, Universitä/t Wien, Wä/hringerstrasse 17, A-1090 Wein, Austria xtof@tbi.univie.ac.at

  • Venue:
  • Genetic Programming and Evolvable Machines
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Fitness landscapes have proved to be a valuable concept in evolutionary biology, combinatorial optimization, and the physics of disordered systems. Usually, a fitness landscape is considered as a mapping from a configuration space equipped with some notion of adjacency, nearness, distance, or accessibility, into the real numbers. In the context of multi-objective optimization problems this concept can be extended to poset-valued landscapes. In a geometric analysis of such a structure, local Pareto points take on the role of local minima. We show that the notion of saddle points, barriers, and basins can be extended to the poset-valued case in a meaningful way and describe an algorithm that efficiently extracts these features from an exhaustive enumeration of a given generalized landscape.