Efficient and effective random testing using the Voronoi diagram

Abstract

Adaptive random testing (ART) is a method for improving the fault-finding effectiveness of random testing. Fixed-size candidate set ART is the most studied variant of this approach. However, existing implementations of FSCS-ART have had substantial selection overhead, with n test cases requiring O(n/sup 2/) time to generate. We describe the use of a geometric data structure known as the Voronoi diagram to reduce this overhead to no worse than O(n/spl radic/n) and, with further optimization, O(nlogn). We demonstrate experimentally that practical improvements in selection overhead can be gained using this improved implementation.