mlsanityData to 7 Oct 2026

Reliability Assurance for Deep Neural Network Architectures Against Numerical Defects

arXiv 2302.06086 ↗First posted 13 Feb 2023cs.SE cs.LG cs.PLcited by 3

Not on the map. Why?

Abstract (arXiv)

With the widespread deployment of deep neural networks (DNNs), ensuring the reliability of DNN-based systems is of great importance. Serious reliability issues such as system failures can be caused by numerical defects, one of the most frequent defects in DNNs. To assure high reliability against numerical defects, in this paper, we propose the RANUM approach including novel techniques for three reliability assurance tasks: detection of potential numerical defects, confirmation of potential-defect feasibility, and suggestion of defect fixes. To the best of our knowledge, RANUM is the first approach that confirms potential-defect feasibility with failure-exhibiting tests and suggests fixes automatically. Extensive experiments on the benchmarks of 63 real-world DNN architectures show that RANUM outperforms state-of-the-art approaches across the three reliability assurance tasks. In addition, when the RANUM-generated fixes are compared with developers' fixes on open-source projects, in 37 out of 40 cases, RANUM-generated fixes are equivalent to or even better than human fixes.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 32 of its 72 references that are in the map. First posted on arXiv 13 Feb 2023.

Builds on

32 of 72 references.

References · 32

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  4. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  5. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  6. 2019Deep Leakage from Gradientscited by 1,004
  7. 2016Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networkscited by 895
  8. 2016TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systemscited by 660
  9. 2020On Adaptive Attacks to Adversarial Example Defensescited by 547
  10. 2017DeepXplore: Automated Whitebox Testing of Deep Learning Systemscited by 307
  11. 2019Machine Learning Testing: Survey, Landscapes and Horizonscited by 213
  12. 2017Bounding and Counting Linear Regions of Deep Neural Networkscited by 161
20 more
  1. 2018DeepGauge: Multi-Granularity Testing Criteria for Deep Learning Systemscited by 144
  2. 2017Convergence Analysis of Two-layer Neural Networks with ReLU Activationcited by 137
  3. 2020SoK: Certified Robustness for Deep Neural Networkscited by 100
  4. 2019Taxonomy of Real Faults in Deep Learning Systemscited by 90
  5. 2021PRIMA: General and Precise Neural Network Certification via Scalable Convex Hull Approximationscited by 77
  6. 2018TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzingcited by 67
  7. 2021Introduction to Neural Network Verificationcited by 63
  8. 2021Provable Repair of Deep Neural Networkscited by 39
  9. 2021DeepLocalize: Fault Localization for Deep Neural Networkscited by 37
  10. 2021RobOT: Robustness-Oriented Testing for Deep Learning Systemscited by 27
  11. 2021Silent Bugs in Deep Learning Frameworks: An Empirical Study of Keras and TensorFlowcited by 25
  12. 2020InferCode: Self-Supervised Learning of Code Representations by Predicting Subtreescited by 23
  13. 2020NeuroDiff: Scalable Differential Verification of Neural Networks using Fine-Grained Approximationcited by 22
  14. 2021Certified Robustness to Programmable Transformations in LSTMscited by 18
  15. 2021Self-Checking Deep Neural Networks in Deploymentcited by 18
  16. 2022DeepStability: A Study of Unstable Numerical Methods and Their Solutions in Deep Learningcited by 11
  17. 2022LinSyn: Synthesizing Tight Linear Bounds for Arbitrary Neural Network Activation Functionscited by 10
  18. 2021Operation is the hardest teacher: estimating DNN accuracy looking for mispredictionscited by 8
  19. 2020Detecting and Understanding Real-World Differential Performance Bugs in Machine Learning Librariescited by 7
  20. 2021ArchRepair: Block-Level Architecture-Oriented Repairing for Deep Neural Networkscited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 1; 2026: 1.

3 papers.

  1. 2025Training with Confidence: Catching Silent Errors in Deep Learning Training with Automated Proactive Checkscited by 10
  2. 2024Terrain Classification Enhanced with Uncertainty for Space Exploration Robots from Proprioceptive Datacited by 0
  3. 2026BTTackler: A Diagnosis-based Framework for Efficient Deep Learning Hyperparameter Optimizationcited by 0