ISSN 2071-8594

Russian academy of sciences

Editor-in-Chief

Gennady Osipov

P. Sh. Geidarov On the Possibility of Determining Values of the Neural Network Weights by an Electrostatic Field

Abstract.

At present, in typical architectures of feedforward neural networks, the values of the weights of connections and thresholds of neurons are determined by adjusting the values of the weights performed by means of typical learning algorithms. Also known are the architectures of feedforward neural networks, implemented on the basis of metric recognition methods, for which the values of the weights of neurons are pre-calculated analytically. The analytical calculation of the val-ues of the weights is carried out on the basis of metric expressions and allows to immediately obtain a workable neural network without training. In this case, the effectiveness of the obtained neural network will depend on the selected set and the number of samples, as well as on the selected dimension of the table of weights. It was also shown that such neural networks can be additionally trained with typical learning algorithms, which makes it possible to increase the efficiency of the neural network with the calculated weights by additional training of the neural network. In this case, the process of calculating the values of the weights and further training of the neural network is also faster than train-ing the neural network in the traditional way, by means of a random initial generation of the weight values of the neural network. In this work, on the basis of these networks, the possibility of determin-ing the values of weights and thresholds of a neural network using the parameters of the electrostatic field: tension, potential is considered. That is, it is proposed to use the values of the parameters of the electrostatic field as the values of the weights of the neural network. In other words, the possibility of creating a workable neural network without analytical calculations and without the use of learning al-gorithms is considered. This approach allows to make the process of determining the values of the neural network weights almost instantaneous. The technical possible implementations of this approach and the problematic aspects of using the parameters of the electrostatic field as the weights of the neural network, as well as possible approaches to resolving these difficulties are considered.

Keywords:

neural networks, electric field, electric field strength, electric potential, learning algorithms, neurocomputer.

PP. 33-49.

DOI 10.14357/20718594210104

References

1. Gorban' A.N., Dudin-Barkovskij V.L., Kirdin A.N. i dr. Nejroinformatika [Neuroinformatics]. Novosibirsk: Nauka. 1998. P. 296.
2. Kruglov V.V., Borisov V.V. Iskusstvennye nejronnye seti. Teorija i praktika [Artificial neural networks. Theory and practice]. M. 2001. P. 328.
3. Yann LeCun, Yoshua Bengio, Geoffrey Hinton Deep learning // Nature. 2015. V. 521. N 7553. P. 436-444.
4. Bazenkov N.I., Voroncov D.D., D'jakonova V.E., Zhiljakova L.Ju., Zakharov I.S., Kuznetsov O.P., Kulivets S.G., Diskretnoe modelirovanie mezhnejronnyh vzaimodejstvij v mul'titransmitternyh setjah [Discrete modeling of inter-neural interactions in multitransmitter networks]. Iskusstvennyj intellekt i prinjatie reshenij [Artificial intelligence and decision making]. 2017. № 2. P. 55-73.
5. Osipov V.Ju., Nikiforov V.V. Vozmozhnosti rekurrentnyh nejronnyh setej s upravljaemymi jelementami po vosstanovleniju potokov kadrov [Possibilities of recurrent neural networks with controlled elements for recovering frame streams]. Informacionno-upravljajushhie sistemy [Information cjntrol systems]. 2019. № 5 (102). P. 10-17.
6. Kuznetsov O.P., Bazenkov N.I., Boldyshev B.A., Zhiljakova L.Ju., Kulivets S.G., Chistopol'skij I.A. Asinhronnaja diskretnaja model' himicheskih vzaimodejstvij v prostyh nejronnyh sistemah [Asynchronous discrete model of chemical interactions in simple neural systems]. Iskusstvennyj intellekt i prinjatie reshenij [Artificial intelligence and decision making] 2018. № 2. P. 3-20.
7. Le T.Ch. Sravnenie nejronnoj seti smas i mnogoslojnoj nejronnoj seti v zadache obnaruzhenija dos-atak [Comparison of the smas neural network and the multilayer neural network in the problem of detecting dos attacks]. Nejrokomp'jutery: razrabotka, primenenie [Neurocomputers: development, application] 2016. № 7. P. 65-69.
8. Golov D.V., Krasovskaja L.V. Nejronnye seti i raspoznanie rukopisnyh cifr na osnove iskusstvennyh nejronnyh setej [Neural networks and handwritten digit recognition based on artificial neural networks]. Issledovanija tehnicheskih nauk [Tehnikal Science Re-search] 2014. № 4 (14). P. 18-20.
9. Drokin I.S. Ob odnom algoritme posledovatel'noj inicializacii vesov glubokih nejronnyh setej i obuchenii ansamblja nejronnyh setej [On one algorithm for sequential initialization of weights of deep neural networks and training an ensemble of neural networks]. Vestnik Sankt-Peterburgskogo universiteta. Prikladnaja matematika. Informatika. Processy upravlenija. [Bulletin of St. Petersburg University. Applied Mathematics. Informatics. Management processes.]. 2016. № 4. P. 66-74.
10. Lukina A.S., Nekrasov M.V., Pakman D.N. Obrabotka telemetricheskoj informacii kosmicheskogo apparata nejronnymi setjami na osnove teorii fil'trov kalmana [Processing of telemetric information of the spacecraft by neural networks based on the theory of kalman filters]. Tendencii razvitija nauki i obrazovanija [ Trends in the development of science and education]. 2016. № 13. P. 43-45.
11. Husainov A.T. Ocenka prognoziruemosti sistemy podderzhanija plastovogo davlenija nejronnymi setjami na neftjanyh mestorozhdenijah [Assessment of predictability of the system for maintaining reservoir pressure by neural networks in oil fields]. Akademicheskij zhurnal Zapadnoj Sibiri [Academic Journal of Western Siberia]. 2016. V. 12. No. 3 (64). P. 48.
12. Bondarko V.M., Bondarko D.V., Solnushkin S.D., Chihman V.N. Modelirovanie rezul'tatov psihofizicheskih jeksperimentov nejronnymi setjami. [Modeling the results of psychophysical experiments with neural networks]. Nejrokomp'jutery: razrabotka, primenenie [Neurocomputers: development, application]. 2018. No. 5. P. 31-33.
13. Peng Shi, Fanbiao Li, Ligang Wu, Neural Network-Based Passive Filtering for Delayed Neutral-Type Semi-Markovian Jump Systems. IEEE Transactions on Neural Networks and Learning Systems. 2017. V. 28. I. 9. P. 210- 2114.
14. Chenggang Yan, Hongtao Xie , Dongbao Yang, Jian Yin, Yongdong Zhang , Qionghai Dai, Supervised Hash Coding With Deep Neural Network for Environment Perception of Intelligent Vehicles. EEE Transactions on Intelligent Transportation Systems. 2018. V. 19. I. 1. P. 284 - 295.
15. Wei He, Yuhao Chen, Zhao Yin, Adaptive Neural Network Control of an Uncertain Robot With Full-State Constraints. IEEE Transactions on Cybernetics. 2016. V. 46. I. 3. P. 620- 629.
16. Andre Esteva, Brett Kuprel, Roberto A. Novoa, Justin Ko, Susan M. Swetter, Helen M. Blau & Sebastian Thrun, Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017. V. 542. P. 115–118.
17. Geidarov, P.Sh. Neural networks on the basis of the sample method. Automatic Control And Computer Sciences. 2009. V. 43. No. 4. P. 203-210.
18. Gejdarov, P.Sh. Arhitektura nejronnoj seti s poparno posledovatel'nym razdeleniem obrazov.[ Neural network architecture with pairwise sequential separation of images]. Prikladnaja diskretnaja matematika [Applied discrete mathematics]. 2018. No.41. P. 98-109.
19. Gejdarov P. Sh. Nejronnye seti na osnove metricheskih metodov raspoznavanija v primenenii k zadacham s nechetkimi vyvodami [Neural networks based on metric recognition methods as applied to problems with fuzzy inference]. Iskusstvennyj intellekt i prinjatija reshenij [Artificial intelligence and decision making]. 2010. No. 2. P. 77-88.
20. Geidarov P. Sh. Multitasking application of neural networks implementing metric methods of recognition, Autom. Remote Control. 2013. V. 74. P. 1474-1485.
21. Geidarov, P.Sh. Clearly Defined Neural Networks Architecture. Optical Memory & Neural Networks. 2015. V. 24. I. 3. P. 209-219.
22. Geidarov P. Sh. Neural Networks with Image Recognition by Pairs. Optical Memory and Neural Networks. 2018. V. 27. No. 2. P. 113–119.
23. Gejdarov P. Sh. Algoritm vychislenija znachenij vesov sinapsov pervogo sloja nejronnoj seti na osnove metricheskih metodov raspoznavanija. Chast' 1 [An algorithm for calculating the values of the weights of the synapses of the first layer of the neural network based on metric recognition methods. Part 1]. Informacionno-Upravljajushhie Sistemy [Information Control Systems]. Sankt-Peterburg. 2020. No.2. P. 20-30.
24. Gejdarov P. Sh. Algoritm vychislenija znachenij vesov sinapsov pervogo sloja nejronnoj seti na osnove metricheskih metodov raspoznavanija. Chast' 2. [An algorithm for calculating the values of the weights of the synapses of the first layer of the neural network based on metric recognition methods. Part 2]. Informacionno-Upravljajushhie Sistemy [Information Control Systems].Sankt-Peterburg. 2020. No.3. P. 25-38.
25. Birger I.A. Tehnicheskaja diagnostika [Technical diagnostics]. M.:Mashinostroenie. 1978.
26. Gejdarov, P.Sh. Algoritm realizacii metoda blizhajshego soseda v mnogoclojnom perseptrone. [Algorithm for implementing the nearest neighbor method in a multilayer perceptron]. Trudy SPIIRAN [ SPIIRAS Proceedings]. 2017. V. 51. P. 123-151.
27. Geidarov P. Sh. Comparative Analysis of the Results of Training a Neural Network with Calculated Weights and with Random Generation of the Weights. Automation and Remote Control. 2020. V. 81. No. 7. P. 1211-1229.
28. Birjukov S.V. Fizicheskie osnovy izmerenija parametrov jelektricheskih polej [Physical foundations of measuring the parameters of electric fields]. Omsk: SibADI. 2008. P. 112.