Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
166 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

UR4NNV: Neural Network Verification, Under-approximation Reachability Works! (2401.12550v1)

Published 23 Jan 2024 in cs.AI and cs.LG

Abstract: Recently, formal verification of deep neural networks (DNNs) has garnered considerable attention, and over-approximation based methods have become popular due to their effectiveness and efficiency. However, these strategies face challenges in addressing the "unknown dilemma" concerning whether the exact output region or the introduced approximation error violates the property in question. To address this, this paper introduces the UR4NNV verification framework, which utilizes under-approximation reachability analysis for DNN verification for the first time. UR4NNV focuses on DNNs with Rectified Linear Unit (ReLU) activations and employs a binary tree branch-based under-approximation algorithm. In each epoch, UR4NNV under-approximates a sub-polytope of the reachable set and verifies this polytope against the given property. Through a trial-and-error approach, UR4NNV effectively falsifies DNN properties while providing confidence levels when reaching verification epoch bounds and failing falsifying properties. Experimental comparisons with existing verification methods demonstrate the effectiveness and efficiency of UR4NNV, significantly reducing the impact of the "unknown dilemma".

Definition Search Book Streamline Icon: https://streamlinehq.com
References (34)
  1. How good are convex hull algorithms? Comput. Geom., 7:265–301, 1997.
  2. Polyhedral computation, volume 48. American Mathematical Soc., 2009.
  3. Primal - dual methods for vertex and facet enumeration. Discret. Comput. Geom., 20(3):333–357, 1998.
  4. Polyhedral representation conversion up to symmetries. ArXiv, abs/math/0702239, 2007.
  5. Abstract interpretation: a unified lattice model for static analysis of programs by construction or approximation of fixpoints. In Proceedings of the 4th ACM SIGACT-SIGPLAN symposium on Principles of programming languages, pages 238–252, 1977.
  6. A machine program for theorem-proving. Commun. ACM, 5(7):394–397, jul 1962.
  7. Ruediger Ehlers. Formal verification of piece-wise linear feed-forward neural networks. In International Symposium on Automated Technology for Verification and Analysis, pages 269–286. Springer, 2017.
  8. Ai2: Safety and robustness certification of neural networks with abstract interpretation. In 2018 IEEE symposium on security and privacy (SP), pages 3–18. IEEE, 2018.
  9. Basic Properties of Convex Polytopes, page 243–270. CRC Press, Inc., USA, 1997.
  10. Reachnn: Reachability analysis of neural-network controlled systems. ACM Transactions on Embedded Computing Systems (TECS), 18(5s):1–22, 2019.
  11. Verifying the safety of autonomous systems with neural network controllers. ACM Transactions on Embedded Computing Systems (TECS), 20(1):1–26, 2020.
  12. Grounding spatio-temporal language with transformers. In Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual, pages 5236–5249, 2021.
  13. Reluplex: An efficient smt solver for verifying deep neural networks. In Computer Aided Verification, pages 97–117, Cham, 2017. Springer International Publishing.
  14. The marabou framework for verification and analysis of deep neural networks. In Computer Aided Verification, pages 443–452, Cham, 2019. Springer International Publishing.
  15. Open-and closed-loop neural network verification using polynomial zonotopes. arXiv preprint arXiv:2207.02715, 2022.
  16. Qualitative and quantitative model checking against recurrent neural networks. Journal of Computer Science and Technology, 2023.
  17. Safety verification for neural networks based on set-boundary analysis. In Theoretical Aspects of Software Engineering, pages 248–267, Cham, 2023. Springer Nature Switzerland.
  18. Verifying relu neural networks from a model checking perspective. J. Comput. Sci. Technol., 35(6):1365–1381, nov 2020.
  19. Verifying relu neural networks from a model checking perspective. J. Comput. Sci. Technol., 35(6):1365–1381, 2020.
  20. Research on automatic emergency steering collision avoidance and stability control of intelligent driving vehicle. Frontiers Robotics AI, 10, 2023.
  21. A survey and comparison of methods for finding all vertices of convex polyhedral sets. Math. Oper. Res., 5(2):167–185, 1980.
  22. On the number of linear regions of deep neural networks. In Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2, NIPS’14, page 2924–2932, Cambridge, MA, USA, 2014. MIT Press.
  23. Acas xu: Integrated collision avoidance and detect and avoid capability for uas. In 2019 IEEE/AIAA 38th Digital Avionics Systems Conference (DASC), pages 1–10, 2019.
  24. Bounding and counting linear regions of deep neural networks. ArXiv, abs/1711.02114, 2017.
  25. Fast and effective robustness certification. In Proceedings of the 32nd International Conference on Neural Information Processing Systems, NIPS’18, page 10825–10836, Red Hook, NY, USA, 2018. Curran Associates Inc.
  26. An abstract domain for certifying neural networks. 3(POPL), jan 2019.
  27. Safety verification of cyber-physical systems with reinforcement learning control. 18(5s), oct 2019.
  28. Star-based reachability analysis of deep neural networks. In Formal Methods – The Next 30 Years, pages 670–686, Cham, 2019. Springer International Publishing.
  29. Nnv: The neural network verification tool for deep neural networks and learning-enabled cyber-physical systems. In Computer Aided Verification, pages 3–17, Cham, 2020. Springer International Publishing.
  30. Efficient formal safety analysis of neural networks. Advances in Neural Information Processing Systems, 31, 2018.
  31. Reachable set computation and safety verification for neural networks with relu activations. CoRR, abs/1712.08163, 2017.
  32. Output reachable set estimation and verification for multilayer neural networks. IEEE transactions on neural networks and learning systems, 29(11):5777–5783, 2018.
  33. Improving neural network verification through spurious region guided refinement. In International Conference on Tools and Algorithms for the Construction and Analysis of Systems, pages 389–408. Springer, 2021.
  34. Bartscore: Evaluating generated text as text generation. In Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual, pages 27263–27277, 2021.

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets