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
  • [31] Improving empirical models with machine learning
    Bhattacharya, Biswa
    Solomatine, Dimitri
    2006 IEEE INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORK PROCEEDINGS, VOLS 1-10, 2006, : 4854 - 4861
  • [32] An Empirical Review of Automated Machine Learning
    Vaccaro, Lorenzo
    Sansonetti, Giuseppe
    Micarelli, Alessandro
    COMPUTERS, 2021, 10 (01) : 1 - 27
  • [33] Machine learning in empirical asset pricing
    Alois Weigand
    Financial Markets and Portfolio Management, 2019, 33 : 93 - 104
  • [34] Machine learning in empirical asset pricing
    Weigand, Alois
    FINANCIAL MARKETS AND PORTFOLIO MANAGEMENT, 2019, 33 (01) : 93 - 104
  • [35] Empirical Forecasting Analysis of Bitcoin Prices: A Comparison of Machine Learning, Deep Learning, and Ensemble Learning Models
    Tripathy, Nrusingha
    Hota, Sarbeswara
    Mishra, Debahuti
    Satapathy, Pranati
    Nayak, Subrat Kumar
    INTERNATIONAL JOURNAL OF ELECTRICAL AND COMPUTER ENGINEERING SYSTEMS, 2024, 15 (01) : 21 - 29
  • [36] Machine learning solutions in sewer systems: a bibliometric analysis
    Ribalta, Marc
    Bejar, Ramon
    Mateu, Carles
    Rubion, Edgar
    URBAN WATER JOURNAL, 2023, 20 (01) : 1 - 14
  • [37] Modeling and Security Analysis of Attacks on Machine Learning Systems
    Singhal, Anoop
    PROCEEDINGS OF THE 10TH ACM INTERNATIONAL WORKSHOP ON SECURITY AND PRIVACY ANALYTICS, IWSPA 2024, 2024, : 1 - 2
  • [38] Analysis of the Application of Machine Learning in Automatic Control Systems
    Sviridov, Alexey
    Bobkov, Vladislav
    Lemza, Anastasia
    Balashov, Alexander
    Bobrikov, Dmitriy
    PROCEEDINGS OF THE 2021 IEEE CONFERENCE OF RUSSIAN YOUNG RESEARCHERS IN ELECTRICAL AND ELECTRONIC ENGINEERING (ELCONRUS), 2021,
  • [39] Machine Learning for Reliability Analysis of Large Scale Systems
    Smirni, Evgenia
    QUANTITATIVE EVALUATION OF SYSTEMS (QEST 2020), 2020, 12289 : 3 - 7
  • [40] Simulation and Analysis for Backward Compatibility of Solder Joints under Thermal Cycle
    Ye-xiang, Ning
    Kai-lin, Pan
    Ni, Li
    2008 INTERNATIONAL CONFERENCE ON ELECTRONIC PACKAGING TECHNOLOGY & HIGH DENSITY PACKAGING, VOLS 1 AND 2, 2008, : 308 - 313