Statistical Debugging for Simulations

• DOI: 10.1145/2699722
• Publication Date: 2015-04-17T22:12:01Z
• Authors: Ross Gore; Paul F. Reynolds Jr.; David Kamensky; Saikou Diallo; Jose Padilla
• Keywords: Pace; Statement (logic)

Predictions from simulations have entered the mainstream of public policy and decision-making practices. Unfortunately, methods for gaining insight into faulty outputs not kept pace. Ideally, an gathering method would automatically identify cause a output explain to simulation developer how correct it. In field software engineering, this challenge has been addressed general-purpose through statistical debuggers. We present two research contributions, elastic predicates many-valued labeling functions , that enable debuggers designed become more effective employing random variates continuous numbers. Elastic address deficiencies existing related numbers, whereas support use variates. When used in combinations, these contributions allow tasked with localizing program statement causing examine 40% fewer statements than leading alternatives. Our evaluation shows maintain their ability reduce number need be examined under imperfect conditions developers experience practice.