Memristive Logic Design of Multifunctional Spiking Neural Network with Unsupervised Learning
- PDF / 1,908,045 Bytes
- 10 Pages / 595.224 x 790.955 pts Page_size
- 54 Downloads / 188 Views
Memristive Logic Design of Multifunctional Spiking Neural Network with Unsupervised Learning N.V. Andreeva1
· E.A. Ryndin1 · M.I. Gerasimova1
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract We report a prospective approach to neural network modeling based on implementation of metal-oxide heterostructures with non-volatile memory behavior and multilevel resistive switching. These structures could be used as artificial synapses in neural networks, providing a modulation of synaptic strength in the range of seven orders of magnitude. Together with leaky integrated-and-fire neurons, it allows to organize a feed-forward spiking neural network with embedded spiketiming-dependent plasticity mechanism at the hardware level. The results of computer simulation demonstrate an ability of reconstructed networks based on electronic multilevel synapses to unsupervised learning and processing of asynchronous stream of spikes. Following the results of computer simulation, the biologically inspired circuit design for artificial neural network with memristive synapses was developed. The idea underlying the circuit design is based on the analogue approach and implies the hybrid CMOS-neurons/memristive synapses network, where operational amplifiers used as elements of current control through the memristive devices. Keywords Neuromorphic · Multilevel resistive switching · Spiking neural networks
1 Introduction Key approaches to biological inspired electronics, so-called neuromorphic systems, imply a concept synthesis of operating principles and simulation algorithms for computational modeling of biological neural networks [1–4], developing communication protocols for asynchronous physiological data streams [5, 6], innovation technologies of electronic devices [7–10], integration of multibit memory logic in circuit solutions of basic neuromorphic modules [11, 12], and comprehensive researches of essential synchronization between neuromorphic hardware with biological systems [13–17]. These aspects are of interest not only to a variety of different applications, but especially to the interaction of artificial neural systems with their environment [18]. Indeed, brain–computer interface (BCI) and implant system hardware implementation is restricted by a conventional computer architecture demanding a parallel processing of the input data streams (according to the N.V. Andreeva
[email protected] 1
Saint Petersburg Electrotechnical University, Saint Petersburg, 197376, Russia
number of CPU cores) to speed up the data processing (treatment) and is extremely inefficient when working with asynchronous large streams of data and in terms of power consumption. To overcome these obstacles in the last years, metaloxide memristive structures have shown great potential in neuromorphic engineering when used as interconnections (electronic analogue of synapses) between artificial neurons based on CMOS (complementary metal oxide semiconductor) design technology [19–22]. Memristive structures demonstrate the ability of t
Data Loading...
