An Empirical Analysis of Backward Compatibility in Machine Learning Systems

被引:18
|
作者
Srivastava, Megha [1 ]
Nushi, Besmira [1 ]
Kamar, Ece [1 ]
Shah, Shital [1 ]
Horvitz, Eric [1 ]
机构
[1] Microsoft Res, Redmond, WA 98052 USA
关键词
D O I
10.1145/3394486.3403379
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In many applications of machine learning (ML), updates are performed with the goal of enhancing model performance. However, current practices for updating models rely solely on isolated, aggregate performance analyses, overlooking important dependencies, expectations, and needs in real-world deployments. We consider how updates, intended to improve ML models, can introduce new errors that can significantly affect downstream systems and users. For example, updates in models used in cloud-based classification services, such as image recognition, can cause unexpected erroneous behavior in systems that make calls to the services. Prior work has shown the importance of "backward compatibility" for maintaining human trust. We study challenges with backward compatibility across different ML architectures and datasets, focusing on common settings including data shifts with structured noise and ML employed in inferential pipelines. Our results show that (i) compatibility issues arise even without data shift due to optimization stochasticity, (ii) training on large-scale noisy datasets often results in significant decreases in backward compatibility even when model accuracy increases, and (iii) distributions of incompatible points align with noise bias, motivating the need for compatibility aware de-noising and robustness methods.
引用
收藏
页码:3272 / 3280
页数:9
相关论文
共 50 条
  • [1] Testing Machine Learning Systems in Industry: An Empirical Study
    Li, Shuyue
    Guo, Jiaqi
    Lou, Jian-Guang
    Fan, Ming
    Liu, Ting
    Zhang, Dongmei
    2022 ACM/IEEE 44TH INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING: SOFTWARE ENGINEERING IN PRACTICE (ICSE-SEIP 2022), 2022, : 263 - 272
  • [2] Empirical analysis of Machine Learning Techniques for context aware Recommender Systems in the environment of IoT
    Sachdeva, Nitin
    Dhir, Renu
    Kumar, Akshi
    INTERNATIONAL CONFERENCE ON ADVANCES IN INFORMATION COMMUNICATION TECHNOLOGY & COMPUTING, 2016, 2016,
  • [3] Channel Estimation Scheme for WLAN Systems with Backward Compatibility
    Kim, Jee-Hoon
    Yu, Heejung
    Lee, Sok-Kyu
    ETRI JOURNAL, 2012, 34 (03) : 450 - 453
  • [4] An Empirical Study of Refactorings and Technical Debt in Machine Learning Systems
    Tang, Yiming
    Khatchadourian, Raffi
    Bagherzadeh, Mehdi
    Singh, Rhia
    Stewart, Ajani
    Raja, Anita
    2021 IEEE/ACM 43RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE 2021), 2021, : 238 - 250
  • [5] Empirical Analysis of Machine Learning Algorithms for Multiclass Prediction
    Ishfaq, Umar
    Shabbir, Danial
    Khan, Jumshaid
    Khan, Hikmat Ullah
    Naseer, Salman
    Irshad, Azeem
    Shafiq, Muhammad
    Hamam, Habib
    Wireless Communications and Mobile Computing, 2022, 2022
  • [6] Empirical Analysis of Machine Learning Algorithms for Multiclass Prediction
    Ishfaq, Umar
    Shabbir, Danial
    Khan, Jumshaid
    Khan, Hikmat Ullah
    Naseer, Salman
    Irshad, Azeem
    Shafiq, Muhammad
    Hamam, Habib
    WIRELESS COMMUNICATIONS & MOBILE COMPUTING, 2022, 2022
  • [7] Empirical Analysis on Cancer Dataset with Machine Learning Algorithms
    Vital, T. PanduRanga
    Krishna, M. Murali
    Narayana, G. V. L.
    Suneel, P.
    Ramarao, P.
    SOFT COMPUTING IN DATA ANALYTICS, SCDA 2018, 2019, 758 : 789 - 801
  • [8] Machine Learning for Fashion Outfit Compatibility
    Nakamura T.
    Kyokai Joho Imeji Zasshi/Journal of the Institute of Image Information and Television Engineers, 2020, 74 (01): : 40 - 43
  • [9] A NOTE ON BACKWARD COMPATIBILITY
    KENDE, M
    ECONOMICS LETTERS, 1994, 45 (03) : 385 - 389
  • [10] A Taxonomy of Software Engineering Challenges for Machine Learning Systems: An Empirical Investigation
    Lwakatare, Lucy Ellen
    Raj, Aiswarya
    Bosch, Jan
    Olsson, Helena Holmstrom
    Crnkovic, Ivica
    AGILE PROCESSES IN SOFTWARE ENGINEERING AND EXTREME PROGRAMMING, XP 2019, 2019, 355 : 227 - 243