Technology for Improving the Quality of Training Artificial Neural Networks in Road Transport Infrastructure Object Management Tasks

M.I. Bulatov, N.V. Eliseeva, V.E. Petrov


The study is conducted in the field of machine learning with a focus on improving the training quality of artificial neural networks in the context of managing road traffic infrastructure objects. Incorrectly chosen training parameters for an artificial neural network can lead to situations in which the global minimum is not reached, which will negatively affect the accuracy of the network. Research is aimed at developing methods that optimize the learning process based on the analysis of changes in weight gradients. This allows you to increase the accuracy and reliability of the neural network in terms of managing transport facilities. The authors propose an algorithm based on tracking the activity of changes in neural network parameters through the analysis of weight gradient variations. This algorithm allows diagnosing the training process and making decisions to adjust parameters aiming at optimizing the neural network training in tasks related to managing road traffic infrastructure. A technology has been developed for the use of artificial neural networks in the management of road transport infrastructure. The algorithm was studied on the CIFAR-10 data set. The developed algorithm is an important tool for improving the quality of training of an artificial neural network in problems of managing road transport infrastructure objects. The ability to analyze and adjust training based on the dynamics of changes in gradients significantly improves the efficiency of the learning process and increases the chances of achieving the required results.

Full Text:

PDF (Russian)


H.Khelifi; A.Belouahri, “The Impact of Big Data Analytics on Traffic Prediction,” In 2022 International Conference on Ad-vanced Aspects of Software Engineering (ICAASE), IEEE, Sep-tember 2022, pp. 1–6.

J.C. Chedjou, K. Kyamakya, “Cellular neural networks based local traffic signals control at a junction/intersection,” In Proceedings of the 1st IFAC Conference on Embedded Systems 2012 (CESCIT-2012), Wurzburg, Germany, 3-5 April, 2012, pp. 81–85.

S. Araghi, A. Khosravi, D. Creighton, “Optimal design of traffic signal controller, using neural networks and fuzzy logic sys-tems,” In Proceedings of the International Joint Conference on Neural Networks 2014 (IJCNN), Beijing, China, 6-11 July, 2014, pp. 42–47.

G.B. Castro, J.C. Martini, A. Hirakawa, “Biologically-inspired neural network for traffic signal control,” In Proc. of 17th International Conference on Intelligent Transportation Systems 2014 (ITSC), Quingdao, China, 8-11 October, 2014, pp. 2144–2149.

W. Rawat, Z. Wang, “Deep convolutional neural networks for image classification: A comprehensive review,” Neural Comp., vol. 29, no. 9, pp. 2352–2449, 2017.

M. Yousef, K. F. Hussain, U. S. Mohammed, “Accurate, data-efficient, unconstrained text recognition with convolutional neural networks,” Pattern Recognit., vol. 108, p. 107482, 2020.

Y. Shen, et al., “Learning semantic representations using convolutional neural networks for web search,” In WWW '14, 2014, pp. 373-374.

P. Yu, X. Yan, “Stock price prediction based on deep neural networks,” Neural Comput Appl, vol. 32, no. 6, pp. 1609-1628, 2020.

N. Qian, “On the momentum term in gradient descent learning algorithms,” Neural Netw., vol. 12, pp. 145–151, 1999.

Y. Nesterov, A method for unconstrained convex minimization problem with the rate of convergence O(1/k2),” Dokl. Akad. Nauk. SSSR, vol. 269, no. 3, p. 543, 1983.

D. Zeiler, ADADELTA: An Adaptive Learning Rate Method. 2012.

P. Kingma, J. Lei, “Adam: a Method for Stochastic Optimiza-tion,” In 3rd International Conference on Learning Representations ICLR, 2015, pp1–13.

S. P. Bingulac, “On the compatibility of adaptive controllers (Published Conference Proceedings style),” In Proc. 4th Annu. Allerton Conf. Circuits and Systems Theory, New York, 1994, pp. 8–16.


  • There are currently no refbacks.

Abava  Кибербезопасность MoNeTec 2024

ISSN: 2307-8162