ID 原文 译文
56338 最后重点反思了后深度学习时代像素级语义理解乃至人工智能领域所面对的机遇与挑战,提出知识的方向指导和数据的优化驱动是未来人工智能发展的重点关注目标. Finally, we focus on the opportunities andchallenges faced by pixel-level semantic understanding, and even artificial intelligence, in the post-deep-learningera and argue that knowledge guidance and data-driven optimization are fundamental objectives for the futuredevelopment of artificial intelligence.
56339 本文意图从像素级语义理解的定义与发展现状出发,延伸出对当前工作的思考以及对整个领域的反思,强调整个领域面临的风险;在介绍像素级语义理解基础认知的同时对相关技术的发展方向和路径进行深入的思考与深度的展望. From the definition and current development of pixel-level semantic under?standing, in this work, we extend the current conceptual approach and reflect on the risks faced by the wholefield, providing an in-depth discussion and consideration of the developmental direction of related technologies,while clarifying the knowledge base of pixel-level semantic understanding.
56340 测试是提高软件可靠性的重要方法. Testing is an important way of improving the reliability of a software product.
56341 消息传递并行程序中存在的不确定通信语句,使得进程执行顺序具有不确定性,这增加了测试该类程序的难度. Uncertain communi?cation statements in the message-passing parallel program can lead to uncertainty in the order of the executionof processes, which increases the difficulty of testing.
56342 鉴于进程执行顺序对目标路径覆盖难易程度的影响,本文研究消息传递并行程序多路径覆盖调度序列排序方法,以提高多路径覆盖测试数据生成的效率. Taking into consideration the influence of an uncertainprocess-execution order on the difficulty of covering target paths, in this paper, we focus on a method for sort?ing scheduling sequences in a message-passing parallel program to enable multi-path coverage and improve theefficiency of generating test data for multi-path coverage.
56343 首先,在每个调度序列下,以每个采样的程序输入执行程序,生成路径覆盖矩阵; To fulfill this task, we first execute a parallel programon each sample in each scheduling sequence to generate a path coverage matrix.
56344 然后,针对每条目标路径,分别计算与路径覆盖矩阵中每条路径的相似度,生成多个路径相似度矩阵; Then, we produce a number ofpath similarity matrixes based on the similarity of each path in the coverage matrix and for each target.
56345 接着,基于这些路径相似度矩阵的特征量,评价调度序列的性能,并依此对调度序列排序; Next,we evaluate each scheduling sequence with respect to the characteristics of the similarity matrixes, based onwhich we sort the scheduling sequences.
56346 最后,基于调度序列排序集,使用随机采样法,生成覆盖所有目标路径的测试数据,并使用缺陷检测平均百分比(average percentage of faults detected, APFD)指标评估调度序列排序集. Finally, we generate test data for multi-path coverage by performing arandom sampling of the sorted scheduling sequence set, which are evaluated using the APFD index. We appliedthe proposed method to nine benchmark parallel programs, and compared the performances of the random andtraditional methods.
56347 将所提方法应用于9个基准并行程序中,并与随机方法和传统方法进行比较.实验结果表明,所提方法对路径覆盖率没有影响,但显著减少了被测程序执行次数和运行时间. The experimental results indicate that the proposed method has no effect on path coverage,but significantly reduces the number of program executions during the test and running time.