Implementation of discrete cosinus transformation in the input block of the memristor neural processor

About the author:

Alexander D. Pisarev, Cand. Sci. (Tech.), Associate Professor, Department of Applied and Technical Physics, School of Natural Sciences, University of Tyumen, Tyumen, Russia; Senior Researcher, Memristive Materials Laboratory, Center for Nature-Inspired Engineering, University of Tyumen, Tyumen, Russia; spcb.doc@utmn.ru, https://orcid.org/0000-0002-5602-3880

Abstract:

This article describes a study on the border of neural network information technologies and memristor nanoelectronics of processors. From the point of view of information technology, artificial and more complex biomorphic neural networks are learning architectures consisting of a large number of simple solvers. The distribution of a large number of simple calculations decreases the performance consumption of even the most powerful standard processor systems. Therefore, in the development of neural networks, the task of creating a neural processor has become particularly urgent.

Neuroprocessor is understood as hardware specifically designed to implement the neural network model in an efficient manner. Major electronics manufacturers (including IBM, Google, Intel, and Huawei) have already joined the neuroprocessor creation race. The tasks of this direction require not only the developed silicon technologies, but also the use of new elements of nanoelectronics, including memristors.

This paper describes the adaptation to hardware of one of the variants of fast discrete cosine transform algorithms, which is a type of Fourier method. The importance of this study lies in the need to solve the problem of entering standard information into the neural processor.

As a hardware, a 3D logical matrix is used, implemented on nano-technological elements of the combined memristor-diode crossbar electronics. This paper presents a method for increasing the filtering rate by applying simple operations that are performed in parallel in logical connected blocks of super-large 3D logic matrix. The speed of such a system can be extremely high and is determined by the time of one clock pulse, limited only by the speed of operation of the inverter elements and the propagation of signals on the tires of the combined memristor-diode crossbar.

References:

1. Gonzalez R. C., Woods R. E. 2005. Digital Image Processing. Translated from English and edited by P. A. Chochia. Moscow: Tekhnosfera. [In Russian] 
2. Pisarev A. D. 2018. “SPICE-modeling of the processes of associative self learning and unconditional discrimination in the logic unit of a neuroprocessor”. Tyumen State University Herald. Physical and Mathematical Modeling. Oil, Gas, Energy, vol. 4, no 3, pp. 132-145. DOI: 10.21684/2411-7978-2018-4-3-132-145  [In Russian]
3. Udovichenko S. Yu., Pisarev A. D., Busygin A. N., Mayevsky O. V. 2018. “Neuroprocessor on the basis of a combined memristor-diode crossbar”. Nanoindustry, no 5 (84), pp. 344-355. DOI: 10.22184/1993-8578.2018.84.5.344.355 [In Russian]
4. Bobylev A. N., Busygin A. N., Pisarev A. D., Udovichenko S. Yu., Filippov V. A. 2017. “Neuromorphic coprocessor prototype based on mixed metal oxide memristors”. International Journal of Nanotechnology, vol. 14, no 7/8, pp. 698-704. DOI: 10.1504/IJNT.2017.083444
5. Levy Y., Bruck J., Cassuto Y., Friedman E. G. et al. 2014. “Logic operations in memory using a memristive Akers array”. Microelectronics Journal, vol. 45, no 11, pp. 1429-1437. DOI: 10.1016/j.mejo.2014.06.006
6. Li C., Hu M., Li Y., Jiang H. et al. 2018. “Analogue signal and image processing with large memristor crossbars”. Nature Electronics, vol. 1, no 1, pp. 52-59. DOI: 10.1038/s41928-017-0002-z
7. Merolla P. A., Arthur J. V., Alvarez-Icaza R., Cassidy A. S., Sawada J., Akopyan F., Jackson B. L., Imam N., Guo C., Nakamura Y., Brezzo B., Vo I., Esser S. K., Appuswamy R., Taba B., Amir A., Flickner M. D., Risk W. P., Manohar R., Modha D. S. 2014. “A million spiking-neuron integrated circuit with a scalable communication network and interface”. Science, vol. 345, no 6197, pp. 668-673. DOI: 10.1126/science.1254642
8. Pisarev A., Busygin A., Udovichenko S., Maevsky O. 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
9. Prezioso M., Mahmoodi M. R., Bayat F. M., Nili H., Kim H., Vincent A. 2018. “Spike-timing-dependent plasticity learning of coincidence detection with passively integrated memristive circuits”. Nature Communications, vol. 9, no 1. P. 5311. DOI: 10.1038/s41467-018-07757-y
10. Udovichenko S., Pisarev A., Busygin A., Maevsky O. 2017. “3D CMOS, memristor nanotechnology for creating logical and memory matrices of neuroprocessor”. Nanoindustry, no 5, pp. 26-34. DOI: 10.22184/1993-8578.2017.76.5.26.34