Algorithms for building and operation modeling of large electrical circuits with memristor-diode crossbars in a biomorphic neuroprocessor

About the authors:

Abdulla H. Ebrahim, Postgraduate Student, Department of Applied and Technical Physics, University of Tyumen; abdulla.ybragim@mail.ru, ORCID: 0000-0002-1709-9882

Sergey Yu. Udovichenko, Dr. Sci. (Phys.-Math.), Professor of Department of Applied and Technical Physics; Head of REC “Nanotechnology”, University of Tyumen; eLibrary AuthorID, ResearcherID, ScopusID, udotgu@mail.ru

Abstract:

The biomorphic neuroprocessor is the hardware implementation of the impulse neural network in which incoming information from a set of numbers is converted into impulses, and outgoing information, on the contrary, from impulses into binary code. For the automatic building of electrical circuits of the input coding and output decoding units in neuroprocessor using ultra-large logic matrices based on a memristor-diode crossbar, appropriate algorithms have been developed. For the subsequent imitation modeling of information processing in these units, as well as in the memory matrix of the neuroprocessor, the algorithm for calculating large electrical circuits containing memristor-diode crossbars has been created. This simulation algorithm is based on the well-known algorithm of Simulation Program with Integrated Circuit Emphasis and includes original mathematical models of the memristor and the selective element of the Zener diode, as well as the algorithm for modeling the resistive switching of the memristor. The results of imitation modeling using the developed algorithms and corresponding programs showed the operability of the constructed electrical circuits of the input unit in the mode of encoding a binary number into a impulse frequency by a population of three neurons and the output unit of a neuroprocessor that decodes the impulses coming from neurons into binary format as well as the operability of the memory matrix under weighting and summing impulses. The created algorithms and programs package based on them can be used to effectively solve the engineering and technical problem of manufacturing a biomorphic neuroprocessor that requires modeling of information processing in individual neuroprocessor units based on large memristor-diode arrays in order to optimize their parameters.

References:

1. Ebrahim A. H. A., Busygin A. N., Udovichenko S. Yu. 2022. “Mathematical modeling of memristor resistive switching based on mass transfer full model of oxygen vacancies and ions”. Tyumen State University Herald. Physical and Mathematical Modeling. Oil, Gas, Energy, vol. 8, no. 2 (30), pp. 198-214. DOI: 10.21684/2411-7978-2022-8-2-198-214
   [In Russian]
2. Ebrahim A. H. A., Udovichenko S. Yu. 2020. “Mathematical modeling of resistive states and dynamic switching of a metal oxide memristor”. Tyumen State University Herald. Physical and Mathematical Modeling. Oil, Gas, Energy, vol. 6, no. 2 (22), pp. 127-144. DOI: 10.21684/2411-7978-2020-6-2-127-144 [In Russian]
3. Udovichenko S. Yu., Pisarev A. D., Busygin A. N., Bobylev A. N. 2021. “Biomorphous neuroprocessor — prototype of a new generation computer being a carrier of artificial intelligence. Part 2”. Nanoindustry, vol. 14, no. 1 (103), pp. 68-80.
   DOI: 10.22184/1993-8578.2021.14.1.68.79 [In Russian]
4. Biolek D., Di Ventra M., Pershin Yu. V. 2013. “Reliable SPICE simulations of memristors, memcapacitors and meminductors”. Radioengineering, vol. 22, no. 4, pp. 945-968. DOI: 10.48550/arXiv.1307.2717
5. Busygin A. N., Ebrahim A. H. A., Pisarev A. D., Udovichenko S. Yu. 2021. “Input device for a biomorphic neuroprocessor based on a memristor-diode crossbar for the pulse coding of information”. Nanobiotechnology Reports, vol. 16, no. 6, pp. 798-803. DOI: 10.1134/S2635167621060069
6. Busygin A. N., Udovichenko S. Yu., Ebrahim A. H. A., Bobylev A. N., Gubin A. A. 2022. “Mathematical model of metal-oxide memristor resistive switching based on full physical model of heat and mass transfer of oxygen vacancies and ions”. Physica Status Solidi (A). DOI: 10.1002/pssa.202200478
7. Ebrahim A. H. A., Udovichenko S. Yu. 2022. “Automatic building of electrical circuits of biomorphic neuroprocessor units and visualization of their numerical simulation”.
   In: Rocha A., Isaeva E. (eds.). Science and Global Challenges of the 21st Century — Science and Technology. Perm Forum 2021. Lecture Notes in Networks and Systems, vol. 342, pp. 16-23. Cham: Springer. DOI: 10.1007/978-3-030-89477-1_2
8. Filippov V. A., Bobylev A. N., Busygin A. N., Pisarev A. D., Udovichenko S. Yu. 2019. “A biomorphic neuron model and principles of designing a neural network with memristor synapses for a biomorphic neuroprocessor”. Neural Computing and Applications, vol. 32, no. 7, pp. 2471-2485. DOI: 10.1007/s00521-019-04383-7
9. Gollisch T., Meister M. 2008. “Rapid neural coding in the retina with relative spike latencies”. Science, vol. 319, no. 5866, pp. 1108-1111. DOI: 10.1126/science.1149639
10. Pisarev A. D., Busygin A. N., Udovichenko S. Yu., Maevsky O. V. 2018. “3D memory matrix based on a composite memristor-diode crossbar for a neuromorphic processor”. Microelectronic Engineering, vol. 198, pp. 1-7. DOI: 10.1016/j.mee.2018.06.008
11. Pisarev A. D., Busygin A. N., Udovichenko S. Yu., Maevsky O. V. 2020. “A biomorphic neuroprocessor based on a composite memristor-diode crossbar”. Microelectronics Journal, vol. 102, art. 104827. DOI: 10.1016/j.mejo.2020.104827
12. Pisarev A. D., Busygin A. N., Bobylev A. N., Gubin A. A., Udovichenko S. Yu. 2021. “Fabrication technology and electrophysical properties of a composite memristor-diode crossbar used as a basis for hardware implementation of a biomorphic neuroprocessor”. Microelectronic Engineering, vol. 236, art. 111471. DOI: 10.1016/j.mee.2020.111471