Output details
11 - Computer Science and Informatics
University of Exeter
Visualisation and ordering of many-objective populations
<20> This paper was presented at CEC, one of the two premier conferences for evolutionary computation. It is concerned with the presentation and interpretation of a set of many-objective solutions to optimisation problems. This is important, as often these are high-dimensional, making it difficult for problem owners to navigate. We introduce a novel graphical visualisation of these sets, and examine a number of ways of characterising them. The work also proves the equivalence of two common characterisations, the average rank and outflow. We present, for the first time in evolutionary computing, methods to seriate objectives to improve heatmap visualisations.