Dynamic Visualisation of Many-Objective Populations

Abstract

There has been an increase in research activity recently regarding the visualisation of many-objective populations. Two of the main drivers for this have been (i) to aid decision makers in comparing and selecting designs returned from a many-objective optimisation run, and (ii) to help in the selection of solutions in interactive optimisation. In both of these situations there is often a dynamic element – populations evolving over time change their relative relationships, and the quality comparison measure itself can be altered, redefining member relations. Here we illustrate how a number of existing visualisations from various domains may be applied to many-objective populations to aid the understanding of population relations using the d3 package. d3 is inherently dynamic, and will automatically respond to any changes in the base document underpinning the visualisation, allowing the visualisation package to 'bolt-on' to any other program that can produce or update the underlying file.