一句话:针对概率编程与机器学习框架的非确定原语所引发的 flaky tests,研究如何检测这类非确定失败。
概率编程系统以及 Pyro、PyMC3、TensorFlow、PyTorch 等机器学习框架提供了可扩展且高效的推断与训练原语,但这些操作具有非确定性。因此,依赖这些框架的应用程序的测试很难编写:同一代码版本上的测试可能非确定地失败,即 flaky tests。该论文关注此类概率与机器学习应用中的 flaky test 检测问题。从标题与摘要看,其目标是识别由框架非确定行为引发的测试不稳定现象;摘要本身主要说明问题背景与挑战,未披露具体检测技术、工具实现或实验评估结果。
Probabilistic programming systems and machine learning frameworks like Pyro, PyMC3, TensorFlow, and PyTorch provide scalable and efficient primitives for inference and training. However, such operations are non-deterministic. Hence, it is challenging for developers to write tests for applications that depend on such frameworks, often resulting in flaky tests – tests which fail non-deterministically when run on the same version of code.
DOI 原文 ·
@inproceedings{paperbot6035,
title = {Detecting flaky tests in probabilistic and machine learning applications},
author = {Saikat Dutta and August Shi and Rutvik Choudhary and Zhekun Zhang and Aryaman Jain and Sasa Misailovic},
booktitle = {ISSTA},
year = {2020},
doi = {10.1145/3395363.3397366}
}