Efficient English Translation Method and Analysis Based on the Hybrid Neural Network

Wang, Chuncheng

doi:10.1155/2021/9985251

Cited by 2 publications

(1 citation statement)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Further performance improvements are made by implementing the solvers on a GPU [45]. The major limitation of these approaches is that recursive splitting leads to an exponential growth of subproblems [4,17], which requires more advanced branch-and-bound approaches [43].…”

mentioning

confidence: 99%

Automatic Abstraction Refinement in Neural Network Verification using Sensitivity Analysis

Ladner

Althoff

2023

Proceedings of the 26th ACM International Conference on Hybrid Systems: Computation and Control

View full text Add to dashboard Cite

The formal verification of neural networks is essential for their application in safety-critical environments. However, the set-based verification of neural networks using linear approximations often obtains overly conservative results, while nonlinear approximations quickly become computationally infeasible in deep neural networks. We address this issue for the first time by automatically balancing between precision and computation time without splitting the propagated set. Our work introduces a novel automatic abstraction refinement approach using sensitivity analysis to iteratively reduce the abstraction error at the neuron level until either the specifications are met or a maximum number of iterations is reached. Our evaluation shows that we can tightly over-approximate the output sets of deep neural networks and that our approach is up to a thousand times faster than a naive approach. We further demonstrate the applicability of our approach in closed-loop settings. CCS CONCEPTS• Computer systems organization → Neural networks; • General and reference → Verification; • Theory of computation → Abstraction.

show abstract

mentioning

confidence: 99%