A Sound Abstraction Method Towards Efficient Neural Networks Verification

被引:0
|
作者
Boudardara, Fateh [1 ]
Boussif, Abderraouf [1 ]
Ghazel, Mohamed [1 ,2 ]
机构
[1] Technol Res Inst Railenium, 180 rue Joseph Louis Lagrange, F-59308 Valenciennes, France
[2] Univ Gustave Eiffel, COSYS ESTAS, 20 rue Elisee Reclus, F-59666 Villeneuve Dascq, France
关键词
Neural network abstraction; model reduction; Neural network verification; Output range computation;
D O I
10.1007/978-3-031-49737-7_6
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
With the increasing application of neural networks (NN) in safety-critical systems, the (formal) verification of NN is becoming more than essential. Although several NN verification techniques have been developed in recent years, these techniques are often limited to small networks and do not scale well to larger NN. The primarily reason for this limitation is the complexity and non-linearity of neural network models. Abstraction and model reduction approaches, that aim to reduce the size of the neural networks and over-approximate their outcomes, have been seen as a promising research direction to help existing verification methods to handle larger models. In this paper, we introduce a model reduction method for neural networks with non-negative activation functions (e.g., ReLU and Sigmoid). The method relies on merging neurons while ensuring that the obtained model (i.e., the abstract model) over-approximates the original one. Concretely, it consists in merging a set of neurons that have positive outgoing weights and substituting it with a single abstract neuron, while ensuring that if a given property holds on the abstract network, it necessarily holds on the original one. In order to assess the efficiency of the approach, we perform an experimental comparison with two existing model reduction methods on the ACAS Xu benchmark. The obtained results show that our approach outperforms both methods in terms of precision and execution time.
引用
收藏
页码:76 / 89
页数:14
相关论文
共 50 条
  • [1] Abstraction and Refinement: Towards Scalable and Exact Verification of Neural Networks
    Liu, Jiaxiang
    Xing, Yunhan
    Shi, Xiaomu
    Song, Fu
    Xu, Zhiwu
    Ming, Zhong
    [J]. ACM TRANSACTIONS ON SOFTWARE ENGINEERING AND METHODOLOGY, 2024, 33 (05)
  • [2] Towards Efficient Verification of Quantized Neural Networks
    Huang, Pei
    Wu, Haoze
    Yang, Yuting
    Daukantas, Ieva
    Wu, Min
    Zhang, Yedi
    Barrett, Clark
    [J]. THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 19, 2024, : 21152 - 21160
  • [3] Efficient verification of neural networks based on neuron branching and LP abstraction
    Zhao, Liang
    Duan, Xinmin
    Yang, Chenglong
    Liu, Yuehao
    Dong, Yansong
    Wang, Xiaobing
    Wang, Wensheng
    [J]. NEUROCOMPUTING, 2024, 596
  • [4] Towards an Efficient Verification Method for Monotonicity Properties of Chemical Reaction Networks
    Gori, Roberta
    Milazzo, Paolo
    Nasti, Lucia
    [J]. PROCEEDINGS OF THE 12TH INTERNATIONAL JOINT CONFERENCE ON BIOMEDICAL ENGINEERING SYSTEMS AND TECHNOLOGIES, VOL 3 (BIOINFORMATICS), 2019, : 250 - 257
  • [5] nnenum: Verification of ReLU Neural Networks with Optimized Abstraction Refinement
    Bak, Stanley
    [J]. NASA FORMAL METHODS (NFM 2021), 2021, 12673 : 19 - 36
  • [6] An Abstraction-Refinement Approach to Verification of Artificial Neural Networks
    Pulina, Luca
    Tacchella, Armando
    [J]. COMPUTER AIDED VERIFICATION, PROCEEDINGS, 2010, 6174 : 243 - 257
  • [7] Towards Interpreting Recurrent Neural Networks through Probabilistic Abstraction
    Dong, Guoliang
    Wang, Jingyi
    Sun, Jun
    Zhang, Yang
    Wang, Xinyu
    Dai, Ting
    Dong, Jin Song
    Wang, Xingen
    [J]. 2020 35TH IEEE/ACM INTERNATIONAL CONFERENCE ON AUTOMATED SOFTWARE ENGINEERING (ASE 2020), 2020, : 499 - 510
  • [8] Towards Safety Verification of Direct Perception Neural Networks
    Cheng, Chili-Hong
    Huang, Chung-Hao
    Brunner, Thomas
    Hashemi, Vahid
    [J]. PROCEEDINGS OF THE 2020 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE 2020), 2020, : 1640 - 1643
  • [9] Efficient verification by exploiting symmetry and abstraction
    Toshima, Hiroshi
    Yoneda, Tomohiro
    [J]. Systems and Computers in Japan, 2001, 32 (14) : 41 - 53
  • [10] Towards a Verification Flow Across Abstraction Levels
    Gonzalez-de-Aledo, Pablo
    Przigoda, Nils
    Wille, Robert
    Drechsler, Rolf
    Sanchez, Pablo
    [J]. IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2017, 36 (03) : 475 - 488