Authors:
Habib Daneshpajouh
and
Nordin Zakaria
Affiliation:
Universiti Teknologi Petronas, Malaysia
Keyword(s):
Genetic Algorithm, Cluster Formation, Evolutionary Process, Search Space Analysis, Evolution Visualization.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
General Data Visualization
;
Glyph-Based Visualization
;
Interactive Visual Interfaces for Visualization
;
Large Data Visualization
;
Visualization Applications
Abstract:
While Genetic Algorithm (GA) is a powerful tool for combinatorial optimization, the vast population of candidate
solutions it typically deploys and algorithm’s intrinsic randomness lead to difficulty in understanding its
search behavior. We discuss in this paper a clustering-based visualization tool for GA that attempts to mediate
this problem. GA population across its entire generations are clustered, and each cluster and its individuals
are mapped to a visual symbol. The tool enables a GA researcher or user to understand better the behavior
of a GA run, specifically the local searches it performs in its global exploration to go from one generation to
another.