The current successful application of deep learning is based on Deep Neural Network (DNN). Robustness can help users obtain the service quality information of neural network (NN) in practical applications, measure the security of NN, and avoid potential security threats. In the existing robust computing research, there is no method that can give the robustness of a NN for a certain input sample in a timely and effective manner in practical applications. Therefore, in this paper, the robustness features are combined to study the stability of NN. This paper firstly describes the stability and robustness evaluation framework of NN, and then studies the stability of NN from three aspects: robust classification, bifurcation threshold and robustness predictor stability. Robustness indicators and network performance are analyzed and corresponding conclusions are drawn.

Edris Asghari. Neural Network Stability Fusing Robust Features. International Journal of Neural Network (2020), Vol. 1, Issue 1: 41-48. https://doi.org/10.38007/NN.2020.010106.

[1] Bondarev N V. Exploratory, Regression, and Neural Network Analysis of the Stability of Cation Coronates in Selected Pure Solvents. Russian Journal of General Chemistry, 2020, 90(10):1906-1920. https://doi.org/10.1134/S107036322010014X

[2] Woods W, Chen J, Teuscher C. Adversarial explanations for understanding image classification decisions and improved neural network robustness. Nature Machine Intelligence, 2019, 1(11):508-516. https://doi.org/10.1038/s42256-019-0104-6

[3] Kiss A, Bischof B. Robustness Testing of Neural Network for Handwritten Digit Recognition. Alkalmazott Matematikai Lapok, 2020, 37(1):87-101. https://doi.org/10.37070/AML.2020.37.1.04

[4] Gazizov M R, Grigorian K A. Of Neural Network Model Robustness Through Generating Invariant to Attributes Embeddings. Russian Digital Libraries Journal, 2020, 23(6):1142-1154. https://doi.org/10.26907/1562-5419-2020-23-6-1142-1154

[5] Ilina O V, Tereshonok M V. Robustness Study of a Deep Convolutional Neural Network for Vehicle Detection in Aerial Imagery. Journal of Communications Technology and Electronics, 2020, 67(2):164-170. https://doi.org/10.1134/S1064226922020048

[6] Nguyen D A, Minh K D, Nguyen M L, et al. A symbolic execution-based method to perform untargeted attack on feed-forward neural networks. Automated Software Engineering, 2020, 29(2):1-29. https://doi.org/10.1007/s10515-022-00345-x

[7] Strisciuglio N,  Lopez-Antequera M, Petkov N. Enhanced robustness of convolutional networks with a push–pull inhibition layer. Neural Computing and Applications, 2020, 32(3):1-15. https://doi.org/10.1007/s00521-020-04751-8

[8] Szilassy P, B Németh, P Gáspár. Design and robustness analysis of autonomous vehicles in intersections. IFAC-PapersOnLine, 2019, 52(8):321-326. https://doi.org/10.1016/j.ifacol.2019.08.090

[9] Kumar R, Srivastava S. Comparative study of neural networks for dynamic nonlinear systems identification. Soft Computing, 2019, 23(4):1-14. https://doi.org/10.1007/s00500-018-3235-5

[10] Yashas S, Raghunathan S, Priyakumar U D. SCONES: Self-Consistent Neural Network for Protein Stability Prediction Upon Mutation. The Journal of Physical Chemistry B, 2020, 125(38):10657-10671. https://doi.org/10.1021/acs.jpcb.1c04913

[11] Sene N. Fractional input stability and its application to neural network. Discrete and Continuous Dynamical Systems - Series S, 2019, 13(3):853-865. https://doi.org/10.3934/dcdss.2020049

[12] Aljarbouh A, Sharam A, Khateeb M. Using Time-Delay Neural Network (DTDNN) for Enhancement Power Systems Dynamic Stability. International Journal of Computer Applications, 2020, 176(12):40-47. https://doi.org/10.5120/ijca2020920080

[13] Gobinath S, Madheswaran M. Deep perceptron neural network with fuzzy PID controller for speed control and stability analysis of BLDC motor. Soft Computing, 2020, 24(13):10161-10180. https://doi.org/10.1007/s00500-019-04532-z

[14] Alade O, Mahmoud M, Shehri D A, et al. Rapid Determination of Emulsion Stability Using Turbidity Measurement Incorporating Artificial Neural Network (ANN): Experimental Validation Using Video/Optical Microscopy and Kinetic Modeling.. ACS omega, 2020, 6(8):5910-5920. https://doi.org/10.1021/acsomega.1c00017

[15] Nguyen H H, Zieger T, Braatz R D, et al. Robust Control Theory Based Stability Certificates for Neural Network Approximated Nonlinear Model Predictive Control. IFAC-PapersOnLine, 2020, 54(6):347-352. https://doi.org/10.1016/j.ifacol.2020.08.568

[16] Martynyuk A, Stamova I, Martynyuk-Chernienko Y A. On the boundedness and Lagrange stability of fractional-like neural network-based quasilinear systems. The European Physical Journal Special Topics, 2020, 231(10):1789-1799. https://doi.org/10.1140/epjs/s11734-022-00447-3

[17] Nguyen H, Shang L, Bui X N, et al. Toward state-of-the-art techniques in predicting and controlling slope stability in open-pit mines based on limit equilibrium analysis, radial basis function neural network, and brainstorm optimization. Acta Geotechnica, 2020, 17(4):1295-1314. https://doi.org/10.1007/s11440-021-01373-9

[18] Jawa B, Li F, Xywa B, et al. Enhanced stability results for generalized neural networks with time-varying delay - ScienceDirect. Journal of the Franklin Institute, 2020, 357(11):6932-6950. https://doi.org/10.1016/j.jfranklin.2020.04.049