paperbot · PL 论文追踪

全部论文 待读 收藏

P1 Compatibility Issues in Deep Learning Systems: Problems and Opportunities

深度学习系统中的兼容性问题:问题与机遇
FSE Proceedings of the 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering2023匹配度 5/10 · 论文聚焦深度学习系统兼容性问题,属于机器学习系统与软件工程/程序语言交叉,能为面向 ML 系统的自动化检测与修复提供问题背景;但未涉及静态分析技术本身、Agentic 程序、浮点程序分析或贝叶斯程序分析,与用户核心研究方向仅为中等相关。引用 14 inbox
Jun Wang, Guanping Xiao, Shuai Zhang, Huashan Lei, Yepang Liu, Yulei Sui
crossref: 2026-09-12dblp: 2026-09-12

AI 速览

一句话:首个系统性实证研究,刻画深度学习系统兼容性问题的类型、成因、修复策略及自动化检测修复工具现状。

深度学习系统由核心程序、Python、第三方库、底层库、开发工具、操作系统和硬件环境等组件构成,组件间不兼容交互会引发严重兼容性问题,影响开发与部署,但现有研究多关注一般性 DL 缺陷,缺乏对兼容性问题特征的系统认识。本文通过分析 Stack Overflow 上 3,072 条帖子中筛选出的 352 个 DL 兼容性问题,首次开展系统性实证研究,归纳其类型、暴露阶段与症状,总结根因和常用修复策略,并调研自动检测与修复工具的研究现状。评估基于大规模问答社区数据的分类统计与工具调研,而非受控实验;结果揭示了当前自动化能力与实际需求之间的差距,有助于研究者和实践者理解 DL 兼容性问题并推动未来工具开发。

问题
深度学习系统组件间的不兼容交互会导致严重的兼容性问题,显著影响开发与部署,但这类问题的类型、根因、修复方式及自动化检测修复现状尚缺乏系统研究。
方法
对 Stack Overflow 上 3,072 条帖子中分类出的 352 个 DL 兼容性问题进行实证分析,归纳类型、表现阶段、症状、根因与修复策略,并调研自动检测与修复工具。
评估
通过大规模问答社区数据的分类统计与工具现状调查进行评估,而非受控实验,揭示自动化检测与修复 DL 兼容性问题的能力差距。
原文摘要(Abstract)

Deep learning (DL) systems are complex component-based systems, which consist of core program (code implementation and data), Python (language and interpreter), third-party libraries, low-level libraries, development tools, OS, and hardware environments. Incompatible interaction between components would cause serious compatibility issues, substantially affecting the development and deployment processes. What types of compatibility issues are frequently exposed in DL systems? What are the root causes of such issues and how do developers fix them? How far are we from automatically detecting and fixing DL compatibility issues? Although there are many existing studies on DL bugs, the characteristics of DL compatibility issues have rarely been systematically studied and the above questions remain largely unexplored. To fill this gap, we conduct the first comprehensive empirical study to characterize compatibility issues in DL systems. Through analyzing 352 DL compatibility issues classified from 3,072 posts in Stack Overflow, we present their types, manifestation stages, and symptoms. We further summarize the root causes and common fixing strategies, and conduct a tool survey on the current research status of automated detection and repair of DL compatibility issues. Our study allows researchers and practitioners to gain a better understanding of DL compatibility issues and can facilitate future tool development.

链接与引用

DOI 原文 · PDF(开放获取) · DBLP

BibTeX
@inproceedings{WangXZLLS23,
  title = {Compatibility Issues in Deep Learning Systems: Problems and Opportunities},
  author = {Jun Wang and Guanping Xiao and Shuai Zhang and Huashan Lei and Yepang Liu and Yulei Sui},
  booktitle = {FSE},
  year = {2023},
  doi = {10.1145/3611643.3616321}
}