P.I. Kravets, V.N. Shimkovich, D.A.Ferens
ABSTRACT
A method and algorithms of the hardware-software implementation of neuron networks on PLIS of FPGA type using VHDL language is proposed in the work. Examples of the hardware-software implementation by PLIS means of artificial neurons and neuron networks with sigmoid activation functions are presented; it is shown that the considerable velocity of data processing is provided and the use of PLIS resource is optimized owing to the developed method and algorithm.
KEYWORDS
neuron networks, activation functions, PLIS, VHDL language.
REFERENCES
1. Yegupov, N.D. (2001), Metody robastnogo, neiro-nechetkogo i adaptivnogo upravleniya [Methods of robust, neuro-illegible and adaptive control], Izdatelstvo MGTU im. N.E. Bauman, Moscow, Russia.
2. Terekhov, V.A. (2002), Neirosetevye sistemy upravleniya: Ucheb.Posobie dlya vuzov [Neuronetwork systems of control: Manual for Higher Educ. Inst.], Vysshaya shkola, Moscow, Russia.
3. Soloviev, V. (2003), Proektirovanie tsifrovykh system na osnove PLIS [Design of digital systems on PLIS basis], Radio i svyaz, Moscow, Russia.
4. Gilgurt, S.Ya. (2006), “Analysis of the use of reconfiguratble computers based on PLIS for implementation of neuron networks”, Modelyuvannya ta informatsiini tekhnologii. Zb. Nauk.Pr. IPME NAN Ukrainy, Iss. 37, pp. 168-174.
5. Sergienko, A.M. (2003), VHDL dlya proektirovaniya vychislitelnykh ustroistv [VHDL for design of computing devices], ChP “Korneichuk”, OOO “TID “DS”, Kiev, Ukraine.
6. Sergienko, A.M. and Simonenko, V.P. (2007), “Mapping of periodic algorithms to programmed logical integral circuits”, Elektronnoe modelirovanie, Vol. 29, no. 2, pp. 49-61.
7. Kravets, P.I., Lukina, T.I., Zherebko, V.A. and Shimkovich, V.N. (2011), “Methods of Hardware and Software Realization of Adaptive Neural Network PID Controller on FPGAChip”, Journal of Automation and Information Sciences, Vol. 43, Iss. 4, pp. 70-77.
8. Kravets, P.I., Lukina, T.I., Shimkovich, V.N. and Òkach, ².². (2012), “Development and studyof the process of estimation of indices of the control MOMO-objects models”, Visnyk NTUU “KPI”, Informatyka, upravlinnya ta computer engineering: Zb. nauk. Prats, no. 57, pp. 144-150.
9. Kravets, P.I and Shimkovich, V.N. (2013), “Method of optimization of weight coefficients of neuron networks by means of genetic algorithm under implementation on programmed logical integral circuits”, Elektronnoe modelirovanie, Vol. 35, no. 3, pp. 65-75.
10. Kravets, P.I., Shimkovich, V.N. and Îmelchenko, P. (2013), “Neuronetwork components of the systems of control of dynamic objects and their hardware-software implementation on FPGA”, Visnyk NTUU “KPI”, Informatyka, upravlinnya ta computer engineering: Zb. nauk. Prats, no. 59, pp. 78-85. 11. Kravets, P.I., Shimkovich, V.N. and Zubenko, G.A. (2012), “Technology of hardware-software implementation of artificial neuron and artificial neuron networks by FPGA method”, Visnyk NTUU “KPI”, Informatyka, upravlinnya ta computer engineering, no. 55, pp. 174-180.
12. Kravets, P.I., Shimkovich, V.N. and Zubenko, G.A. (2014), “Models of artificial neuron networks under their hardware-software implementation on FPGA”, Intellectualnyi analiz informatsii IAI-2014. Sbornik trudov XIV mezhdunarodnoi nauchnoi konferentsii im. T.A. Taran [Intellectual analysis of information. Collection of proceedings of T.A. Taran ÕIV International Conference], Kiev, May 14-16, 2014, pp. 127-131.