Repairing DNN Architecture: Are We There Yet?

Jinhan Kim*, Nargiz Humbatova, Gunel Jahangirova, Paolo Tonella, Shin Yoo

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review

11 Citations (Scopus)

Abstract

As Deep Neural Networks (DNNs) are rapidly being adopted within large software systems, software developers are increasingly required to design, train, and deploy such models into the systems they develop. Consequently, testing and improving the robustness of these models have received a lot of attention lately. However, relatively little effort has been made to address the difficulties developers experience when designing and training such models: if the evaluation of a model shows poor performance after the initial training, what should the developer change? We survey and evaluate existing state-of-the-art techniques that can be used to repair model performance, using a benchmark of both real-world mistakes developers made while designing DNN models and artificial faulty models generated by mutating the model code. The empirical evaluation shows that random baseline is comparable with or sometimes outperforms existing state-of-the-art techniques. However, for larger and more complicated models, all repair techniques fail to find fixes. Our findings call for further research to develop more sophisticated techniques for Deep Learning repair.

Original languageEnglish
Title of host publication2023 IEEE Conference on Software Testing, Verification and Validation (ICST)
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages234-245
Number of pages12
ISBN (Electronic)9781665456661
ISBN (Print)9781665456678
DOIs
Publication statusE-pub ahead of print - 26 May 2023
Event16th IEEE International Conference on Software Testing, Verification and Validation, ICST 2023 - Dublin, Ireland
Duration: 16 Apr 202320 Apr 2023

Publication series

NameInternational Conference on Software Testing, Verification, and Validation (ICST)
PublisherIEEE

Conference

Conference16th IEEE International Conference on Software Testing, Verification and Validation, ICST 2023
Country/TerritoryIreland
CityDublin
Period16/04/202320/04/2023

Keywords

  • deep learning
  • hyperparameter tuning
  • program repair
  • real faults

Fingerprint

Dive into the research topics of 'Repairing DNN Architecture: Are We There Yet?'. Together they form a unique fingerprint.

Cite this