2020 · 102 citations · 29 references
ComputingComputer EngineeringComputer ScienceComputer VisionMachine LearningSoftware EngineeringDeep LearningPattern RecognitionSoftware MaintenanceAutomated Program RepairBug FixingBug LocalizationPatternsProgram Repair
Deep neural networks are widely used, creating a need for engineering support, especially for repairing DNN‑based software, yet the challenges and repair patterns remain poorly understood. This study aims to identify the challenges and prioritize repair patterns for automated bug repair tools in DNN software. The authors analyzed 970 repair cases from Stack Overflow and GitHub across five major deep learning libraries and released a benchmark of 667 DNN bug–repair pairs. DNN bug fixes differ from traditional patterns, most often correcting data dimensions and network connectivity, but they can introduce adversarial vulnerabilities, new bugs, and pose challenges in localization, model reuse, and rapid release cycles.
Significant interest in applying Deep Neural Network (DNN) has fueled the need to support engineering of software that uses DNNs. Repairing software that uses DNNs is one such unmistakable SE need where automated tools could be beneficial; however, we do not fully understand challenges to repairing and patterns that are utilized when manually repairing DNNs. What challenges should automated repair tools address? What are the repair patterns whose automation could help developers? Which repair patterns should be assigned a higher priority for building automated bug repair tools? This work presents a comprehensive study of bug fix patterns to address these questions. We have studied 415 repairs from Stack Overflow and 555 repairs from GitHub for five popular deep learning libraries Caffe, Keras, Tensorflow, Theano, and Torch to understand challenges in repairs and bug repair patterns. Our key findings reveal that DNN bug fix patterns are distinctive compared to traditional bug fix patterns; the most common bug fix patterns are fixing data dimension and neural network connectivity; DNN bug fixes have the potential to introduce adversarial vulnerabilities; DNN bug fixes frequently introduce new bugs; and DNN bug localization, reuse of trained model, and coping with frequent releases are major challenges faced by developers when fixing bugs. We also contribute a benchmark of 667 DNN (bug, repair) instances.
29
Laurens van der Maaten, Geoffrey E. Hinton · Journal of Machine Learning Research · 2008 · 35.7K citations
Dropout: a simple way to prevent neural networks from overfitting
Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky et al. · 2014 · 34.3K citations
Understanding interobserver agreement: the kappa statistic.
Anthony J. Viera, Joanne M. Garrett · PubMed · 2005 · 7.1K citations
Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
Xinyun Chen, Chang Liu, Bo Li et al. · arXiv (Cornell University) · 2017 · 1K citations · Full text
Shan Lu, Soyeon Park, Eunsoo Seo et al. · 2008 · 892 citations