Corpus ID: 232109506

DeepLocalize: Fault Localization for Deep Neural Networks

@article{Wardat2021DeepLocalizeFL,
  title={DeepLocalize: Fault Localization for Deep Neural Networks},
  author={Mohammad Wardat and Wei Le and H. Rajan},
  journal={ArXiv},
  year={2021},
  volume={abs/2103.03376}
}
Deep neural networks (DNNs) are becoming an integral part of most software systems. Previous work has shown that DNNs have bugs. Unfortunately, existing debugging techniques don’t support localizing DNN bugs because of the lack of understanding of model behaviors. The entire DNN model appears as a black box. To address these problems, we propose an approach and a tool that automatically determines whether the model is buggy or not, and identifies the root causes for DNN errors. Our key insight… Expand

References

SHOWING 1-10 OF 64 REFERENCES
Repairing Deep Neural Networks: Fix Patterns and Challenges
  • 12
  • PDF
MODE: automated neural network model debugging via state differential analysis and input selection
  • 64
  • PDF
DeepFault: Fault Localization for Deep Neural Networks
  • 19
  • PDF
A comprehensive study on deep learning bug characteristics
  • 40
  • PDF
On decomposing a deep neural network into modules
  • 2
  • PDF
Symbolic Execution for Attribution and Attack Synthesis in Neural Networks
  • 10
  • PDF
Symbolic Execution for Importance Analysis and Adversarial Generation in Neural Networks
  • 5
  • PDF
An Empirical Study of Common Challenges in Developing Deep Learning Applications
  • 31
  • PDF
An empirical study on TensorFlow program bugs
  • 75
  • PDF
SemFix: Program repair via semantic analysis
  • 471
  • PDF
...
1
2
3
4
5
...