eLibrary >
PD CEN/CLC ISO/IEC/TR 24029-1:2023 Artificial Intelligence (AI). Assessment of the robustness of neural networks - Overview >
PD CEN/CLC ISO/IEC/TR 24029-1:2023 Artificial Intelligence (AI). Assessment of the robustness of neural networks - Overview, 2024
- undefined
- Foreword
- Introduction
- 1 Scope
- 2 Normative references
- 3 Terms and definitions
- 4 Overview of the existing methods to assess the robustness of neural networks
- 4.1 General
- 4.1.1 Robustness concept
- 4.1.2 Typical workflow to assess robustness
- 4.2 Classification of methods
- 5 Statistical methods
- 5.1 General
- 5.2 Robustness metrics available using statistical methods
- 5.2.1 General
- 5.2.2 Examples of performance measures for interpolation
- 5.2.3 Examples of performance measures for classification
- 5.2.4 Other measures
- 5.3 Statistical methods to measure robustness of a neural network
- 5.3.1 General
- 5.3.2 Contrastive measures
- 6 Formal methods
- 6.1 General
- 6.2 Robustness goal achievable using formal methods
- 6.2.1 General
- 6.2.2 Interpolation stability
- 6.2.3 Maximum stable space for perturbation resistance
- 6.3 Conduct the testing using formal methods
- 6.3.1 Using uncertainty analysis to prove interpolation stability
- 6.3.2 Using solver to prove a maximum stable space property
- 6.3.3 Using optimization techniques to prove a maximum stable space property
- 6.3.4 Using abstract interpretation to prove a maximum stable space property
- 7 Empirical methods
- 7.1 General
- 7.2 Field trials
- 7.3 A posteriori testing
- 7.4 Benchmarking of neural networks
- Annex A (informative) Data perturbation
- Annex B (informative) Principle of abstract interpretation
- Bibliography [Go to Page]
