基于重要抽样的软件统计测试加速

本文提出一种基于重要抽样的软件统计测试加速方法，该方法通过调整软件Markov链使用模型的迁移概率，在根据统计测试结果得到软件可靠性无偏估计的前提下，可以有效提高安全攸关软件的测试效率，部分解决了安全攸关软件统计测试时间和费用开销过大的问题。同时，本文给出了计算优化迁移概率的模拟退火算法。实验仿真结果表明，该方法可以有效地提高安全攸关软件统计测试的效率。

Software Statistical Test Acceleration Based on Importance Sampling

The Markov chain usage model based statistical testing for safety critical software often costs too much time, thus makes the software reliability estimation very difficult. This paper discusses software statistical testing acceleration based on the importance sampling technique, which is one of the classical techniques for improving the efficiency of Monte Carlo simulation. By adjusting the transition probabilities in Markov chain usage models,we can greatly improve the efficiency of software statistical testing while the unbiased reliability estimates can still be computed. The simulated annealing algorithm for computing optimal transition probabilities of the Markov chain usage model is also presented. A case study shows the method can effectively accelerate software statistical testing.