Neuromorphic elements and systems as the basis for the physical implementation of artificial intelligence technologies
- PDF / 845,318 Bytes
- 10 Pages / 612 x 792 pts (letter) Page_size
- 71 Downloads / 158 Views
ACE AND THIN FILMS
Neuromorphic Elements and Systems As the Basis for the Physical Implementation of Artificial Intelligence Technologies V. A. Demina,b,*, A. V. Emelyanova,b,**, D. A. Lapkina,b, V. V. Erokhina,c, P. K. Kashkarova,b,d,e, and M. V. Kovalchuk a,b,d,e a National
Research Centre “Kurchatov Institute”, Moscow, 123182 Russia Moscow Institute of Physics and Technology (State University), Dolgoprudnyi, Moscow oblast, 141700 Russia c CNR-IMEM (National Research Council, Institute of Materials for Electronics and Magnetism) and University of Parma, 42124 Parma, Italy d Faculty of Physics, Moscow State University, Moscow, 119991 Russia e Faculty of Physics, St. Petersburg State University, St. Petersburg, 199034 Russia * e-mail: [email protected] ** e-mail: [email protected] b
Received October 21, 2015
Abstract—The instrumental realization of neuromorphic systems may form the basis of a radically new social and economic setup, redistributing roles between humans and complex technical aggregates. The basic elements of any neuromorphic system are neurons and synapses. New memristive elements based on both organic (polymer) and inorganic materials have been formed, and the possibilities of instrumental implementation of very simple neuromorphic systems with different architectures on the basis of these elements have been demonstrated. DOI: 10.1134/S1063774516060067
INTRODUCTION Against the background of the intense and constantly increasing activity in the study of individual subsystems of the higher nervous system and the living brain as a whole, an urgent problem is to design neuromorphic elements and networks on their basis from artificial inorganic and organic materials and to study their properties. Attempts in this field have been made for a rather long time (since the end of the 1950s), but only the development of a high-power interdisciplinary scientific and technological basis for the instrumental implementation of neuromorphic systems (NSs)—nanotechnologies—leads to an explosive rise in the number of the corresponding studies. Currently, the problem of simulating some simple functions of the human brain is being successfully solved by program tools through the formation of artificial neural networks. The accumulated experience and success in the use of artificial neural networks in different fields of human activity (pattern recognition (e.g., recognition of texts, images, speech, music, and spam), trading on the exchange, use of neural networks as personal assistants, robotics, etc.) have made them prototypes of instrumental NSs. The physical implementation of artificial neural networks is attractive for the following reason. The instrumental implementation of superparallel systems (composed of sev-
eral millions of individual elements with functions of simultaneous data storage and processing, which perform calculations practically independently) on modern computers with successive (von Neumann) architecture (the processor and memory are physically spaced and the operating speed is
Data Loading...
