Abstract

Abstract

Abstract. In fast-paced software projects, engineers don't have the time or the resources to build heavyweight complete descriptions of their software. The best they can do is lightweight incomplete descriptions which may contain missing and contradictory information. Reasoning about incomplete and contradictory knowledge is notoriously difficult. However, recent results from the empirical AI community suggest that randomized search can tame this difficult problem. In this article we demonstrate the the relevance and the predictability of randomized search for reasoning about lightweight models.

No abstract