Efficient Verification of Neural Networks against LVM-based Specifications

被引:0
|
作者
Hanspal, Harleen [1 ,2 ]
Lomuscio, Alessi [2 ]
机构
[1] Imperial Coll London, London, England
[2] Safe Intelligence, London, England
关键词
D O I
10.1109/CVPR52729.2023.00379
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The deployment of perception systems based on neural networks in safety critical applications requires assurance on their robustness. Deterministic guarantees on network robustness require formal verification. Standard approaches for verifying robustness analyse invariance to analytically defined transformations, but not the diverse and ubiquitous changes involving object pose, scene viewpoint, occlusions, etc. To this end, we present an efficient approach for verifying specifications definable using Latent Variable Models that capture such diverse changes. The approach involves adding an invertible encoding head to the network to be verified, enabling the verification of latent space sets with minimal reconstruction overhead. We report verification experiments for three classes of proposed latent space specifications, each capturing different types of realistic input variations. Differently from previous work in this area, the proposed approach is relatively independent of input dimensionality and scales to a broad class of deep networks and real-world datasets by mitigating the inefficiency and decoder expressivity dependence in the present state-of-the-art.
引用
下载
收藏
页码:3894 / 3903
页数:10
相关论文
共 50 条
  • [31] Practical verification of multi-agent systems against SLK specifications
    Cermak, Petr
    Lomuscio, Alessio
    Mogavero, Fabio
    Murano, Aniello
    INFORMATION AND COMPUTATION, 2018, 261 : 588 - 614
  • [32] Neuron importance based verification of neural networks via divide and conquer
    Dong, Yansong
    Liu, Yuehao
    Zhao, Liang
    Tian, Cong
    Duan, Zhenhua
    NEUROCOMPUTING, 2024, 565
  • [33] Verification of a high-speed machining model based on neural networks
    Kaldos, A
    Boyle, A
    Schulz, G
    ADVANCES IN MANUFACTURING TECHNOLOGY - XIII, 1999, : 45 - 49
  • [34] Continuous Safety Verification of Neural Networks
    Cheng, Chih-Hong
    Yan, Rongjie
    PROCEEDINGS OF THE 2021 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE 2021), 2021, : 1478 - 1483
  • [35] Scalable Verification of Quantized Neural Networks
    Henzinger, Thomas A.
    Lechner, Mathias
    Zikelic, Dorde
    THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2021, 35 : 3787 - 3795
  • [36] Model Based Verification of Spiking Neural Networks in Cyber Physical Systems
    Pradhan, Ankit
    King, Jonathan
    Pinisetty, Srinivas
    Roop, Partha S. S.
    IEEE TRANSACTIONS ON COMPUTERS, 2023, 72 (09) : 2426 - 2439
  • [37] Towards Formal Verification of Neural Networks: A Temporal Logic Based Framework
    Wang, Xiaobing
    Yang, Kun
    Wang, Yanmei
    Zhao, Liang
    Shu, Xinfeng
    STRUCTURED OBJECT-ORIENTED FORMAL LANGUAGE AND METHOD (SOFL+MSVL 2019), 2020, 12028 : 73 - 87
  • [38] An Efficient FIFO Based Accelerator for Convolutional Neural Networks
    Panchbhaiyye, Vineet
    Ogunfunmi, Tokunbo
    JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY, 2021, 93 (10): : 1117 - 1129
  • [39] Finger-Vein Verification Based on LSTM Recurrent Neural Networks
    Qin, Huafeng
    Wang, Peng
    APPLIED SCIENCES-BASEL, 2019, 9 (08):
  • [40] Advances in verification of ReLU neural networks
    Ansgar Rössig
    Milena Petkovic
    Journal of Global Optimization, 2021, 81 : 109 - 152