scholarly journals Биоморфный нейропроцессор – прототип компьютера нового поколения, являющегося носителем искусственного интеллекта. Часть 2

2021 ◽  
Vol 14 (1) ◽  
pp. 68-79
Author(s):  
С.Ю. Удовиченко ◽  
А.Д. Писарев ◽  
А.Н. Бусыгин ◽  
А.Н. Бобылев
Keyword(s):  
Neural Network ◽  
Binary Code ◽  
Digital Information ◽  
Electrical Model ◽  
Input And Output ◽  
New Knowledge ◽  
New Association ◽  
Hardware Neural Network ◽  
First Time ◽  
Self Learning

Во входном и выходном устройствах биоморфного нейропроцессора происходят первичная и конечная обработка информации. Представлены результаты по сжатию на входе цифровой информации и ее кодированию в импульсы, а также по декодированию информации об активации нейронов на выходе в цифровой двоичный код. Представлена реализация аппаратной нейросети процессора на основе оригинальной биоморфной электрической модели нейрона. Приведены результаты SPICE-моделирования и экспериментального исследования процессов обработки сигналов в режимах маршрутизации выходных импульсов нейронов на синапсы других нейронов в логической матрице, скалярного умножения матрицы чисел на вектор, а также ассоциативного самообучения в запоминающей матрице. Впервые продемонстрирована генерация новой ассоциации (нового знания) как в компьютерном моделировании, так и в изготовленном мемристорно-диодном кроссбаре, в отличие от самообучения в существующих аппаратных нейросетях с синапсами на базе дискретных мемристоров. Primary and ultimate information processing takes place in the input and output devices of the biomorphic neuroprocessor. The results are presented on the compression of digital information at the input and its coding into pulses, as well as on the decoding of information about the activation of neurons at the output into a digital binary code. An implementation of a hardware neural network of a processor based on an original biomorphic electrical model of a neuron is presented. The results of SPICE modeling and experimental research of signal processing processes in the modes of routing neuron output pulses to synapses of other neurons in a logical matrix, scalar multiplication of a matrix of numbers by a vector, and associative selflearning in a memory matrix are presented. For the first time, the generation of a new association (new knowledge) was demonstrated both in computer simulation and in a fabricated memristor-diode crossbar, in contrast to self-learning in existing hardware neural networks with synapses based on discrete memristors.

THE STUDY OF THE ELECTROPHYSICAL PROPERTIES OF A COMPOSITE MEMRISTOR-DIODE CROSSBAR AS A BASIS OF THE NEUROPROCESSOR HARDWARE IMPLEMENTATION

2020 ◽  
Vol 6 (3) ◽  
pp. 93-109
Author(s):  
Alexander D. Pisarev ◽  
Alexander N. Busygin ◽  
Andrey N. Bobylev ◽  
Alexey A. Gubin ◽  
Sergey Yu. Udovichenko
Keyword(s):  
Signal Processing ◽  
Electrophysical Properties ◽  
Discrete Elements ◽  
Voltage Bias ◽  
Current Voltage ◽  
Positive Voltage ◽  
New Association ◽  
Parasitic Currents ◽  
First Time ◽  
Self Learning

The aim of this article lies in checking the efficiency of memory and logic matrices. Achieving this has required producing a composite memristor-diode crossbar and studying its electrophysical properties. For these purposes, the authors have made a measuring bench, which consists of a composite memristor-diode crossbar, control peripheral circuitry, based on discrete elements with CMOS logic, and Keithley SourceMeter 2400. The silicon junction p-Si/n-Si has been chosen because its electrical properties better suit the Zenner diode’s requirements compared to the p-Si/ZnO junction. The memristor-diode crossbar with the TiN/Ti0,93Al0,07Ox/p-Si/n-Si/W structure was made with implementation of a new diode. The results show that the crossbar cell with a p-Si/n-Si diode has better rectifying properties in comparison with a p-Si/ZnOx diode, because the current in the crossbar cell with positive voltage bias is much higher than with negative voltage bias. Strong rectifying properties of the cell are necessary for the functioning of diode logic in the logic matrix and for memristor state recording in the logic and memory matrices. The study of electrophysical properties of the composite memristor-diode crossbar, measurement of current-voltage characteristics of the diode and composite memristor-diode crossbar cell and signal processing were performed. The signal processing was performed in the following modes: addition of output impulses of neurons and their routing to synapses of other neurons; multiplication of number matrix by vector, performed in the memory matrix with weighing and totalling of signals; and associative self-learning. For the first time, the generation of a new association (new knowledge) in the composite memristor-diode crossbar has been shown, as opposed to associative self-learning in existing hardware neural networks with discrete-memristors-based synapses. The change of crossbar cell’s output current caused by parasitic currents through adjacent cells has been determined. The results show that the control over Zenner diode characteristics allows reducing the power consumption of the composite crossbar. Obtained electrophysical characteristics prove the efficiency of the composite memristor-diode crossbar, intended for production of the memory and logic matrices.

Estimating Pigment Concentrations from Spectral Images Using an Encoder‐Decoder Neural Network

2020 ◽  
Vol 64 (3) ◽  
pp. 30502-1-30502-15
Author(s):  
Kensuke Fukumoto ◽  
Norimichi Tsumura ◽  
Roy Berns
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Absorption Coefficient ◽  
Spectral Data ◽  
High Accuracy ◽  
Pigment Concentration ◽  
Scattering Coefficient ◽  
A Value ◽  
Input And Output ◽  
Pigment Concentrations

Abstract A method is proposed to estimate the concentration of pigments mixed in a painting, using the encoder‐decoder model of neural networks. The model is trained to output a value that is the same as its input, and its middle output extracts a certain feature as compressed information about the input. In this instance, the input and output are spectral data of a painting. The model is trained with pigment concentration as the middle output. A dataset containing the scattering coefficient and absorption coefficient of each of 19 pigments was used. The Kubelka‐Munk theory was applied to the coefficients to obtain many patterns of synthetic spectral data, which were used for training. The proposed method was tested using spectral images of 33 paintings, which showed that the method estimates, with high accuracy, the concentrations that have a similar spectrum of the target pigments.

scholarly journals A Convex Combination Approach for Artificial Neural Network of Interval Data

2021 ◽  
Vol 11 (9) ◽  
pp. 3997
Author(s):  
Woraphon Yamaka ◽  
Rungrapee Phadkantha ◽  
Paravee Maneejuk
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Convex Combination ◽  
Reference Points ◽  
Combination Method ◽  
Suitable Alternative ◽  
Input And Output ◽  
Artificial Neural ◽  
Conventional Models ◽  
Interval Valued

As the conventional models for time series forecasting often use single-valued data (e.g., closing daily price data or the end of the day data), a large amount of information during the day is neglected. Traditionally, the fixed reference points from intervals, such as midpoints, ranges, and lower and upper bounds, are generally considered to build the models. However, as different datasets provide different information in intervals and may exhibit nonlinear behavior, conventional models cannot be effectively implemented and may not be guaranteed to provide accurate results. To address these problems, we propose the artificial neural network with convex combination (ANN-CC) model for interval-valued data. The convex combination method provides a flexible way to explore the best reference points from both input and output variables. These reference points were then used to build the nonlinear ANN model. Both simulation and real application studies are conducted to evaluate the accuracy of the proposed forecasting ANN-CC model. Our model was also compared with traditional linear regression forecasting (information-theoretic method, parametrized approach center and range) and conventional ANN models for interval-valued data prediction (regularized ANN-LU and ANN-Center). The simulation results show that the proposed ANN-CC model is a suitable alternative to interval-valued data forecasting because it provides the lowest forecasting error in both linear and nonlinear relationships between the input and output data. Furthermore, empirical results on two datasets also confirmed that the proposed ANN-CC model outperformed the conventional models.

Study for Workpiece Identity Based on the BP Neural Network and the Zernike Moment

2013 ◽  
Vol 694-697 ◽  
pp. 1958-1963 ◽  
Author(s):  
Xian Wei ◽  
Jing Dong Zhang ◽  
Xue Mei Qi
Keyword(s):  
Neural Network ◽  
Fault Tolerance ◽  
Parallel Processing ◽  
Bp Neural Network ◽  
Complex Shape ◽  
Learning Ability ◽  
Zernike Moment ◽  
Recognition Ability ◽  
Identity Based ◽  
Self Learning

The robots identify, locate and install the workpiece in FMS system by identifying the characteristic information of target workpiece. The paper studied the recognition technology of complex shape workpiece with combination of BP neural network and Zernike moment. The strong recognition ability of Zernike moment can extract the characteristic. The good fault tolerance, classification, parallel processing and self-learning ability of BP neural network can greatly improve the accurate rate of recognition. Experimental results show the effectiveness of the proposed method.

THE CONCEPT OF MANAGING THE DATA NETWORK BASED ON SELF-LEARNING SYSTEMS

2020 ◽  
Author(s):  
С.Р. РОМАНОВ
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Data Transmission ◽  
Network Control ◽  
Learning Systems ◽  
Optimal Routing ◽  
Transmission Network ◽  
Data Traffic ◽  
Data Network ◽  
Self Learning

Рассмотрен принцип управления сетью передачи данных (СПД)с помощью искусственной нейронной сети. Предложена концепция проведения вычислений при решении задачи оптимальной маршрутизации трафика данных. Приведен алгоритм управления сетью СПД на базе нейронной сети Хэмминга. The principle of data transmission network control using an artificial neural network is considered. The concept of carrying out calculations when solving the problem of optimal routing of data traffic is proposed. The algorithm for controlling the data transmission network based on the Hamming neural network is presented.

The Application of Neural Network in Computer Control

2013 ◽  
Vol 380-384 ◽  
pp. 421-424
Author(s):  
Jing Liu ◽  
Yu Chi Zhao ◽  
Xiao Hua Shi ◽  
Su Juan Liu
Keyword(s):  
Neural Network ◽  
Distributed Storage ◽  
Computer Control ◽  
System Optimization ◽  
Good Choice ◽  
Information Storage ◽  
Knowledge Process ◽  
Control Computer ◽  
Self Learning ◽  
Logic System

In recent years, it is a very active direction of research to use neural network to control computer. Neural network is a burgeoning crossing subject, and the way it processes information is different from the past symbolic logic system, which has some unique properties: such as the distributed storage and parallel processing of information, the unity of the information storage and information processing, and have the ability of self-organizing and self-learning. And it has been applied widespread in pattern recognition, signal processing, knowledge process, expert system, optimization, intelligent control and so on. Using neural network can deal with some problems such as complicated environment information, fuzzy background knowledge and undefined inference rules, and it allows samples to have relatively large defects and distortion, so it is a very good choice to adopt the recognizing method of neural network. This thesis discusses the application of neural network in computer control.

scholarly journals High-density information storage in an absolutely defined aperiodic sequence of monodisperse copolyester

2020 ◽  
Vol 11 (1) ◽  
Author(s):  
Jung Min Lee ◽  
Mo Beom Koo ◽  
Seul Woo Lee ◽  
Heelim Lee ◽  
Junho Kwon ◽  
...  
Keyword(s):  
Binary Code ◽  
Storage Capacity ◽  
Information Storage ◽  
Polymer Chemistry ◽  
Digital Information ◽  
Hydroxy Acids ◽  
Storage Density ◽  
Molecular Weights ◽  
Convergent Method ◽  
Scalable Synthesis

AbstractSynthesis of a polymer composed of a large discrete number of chemically distinct monomers in an absolutely defined aperiodic sequence remains a challenge in polymer chemistry. The synthesis has largely been limited to oligomers having a limited number of repeating units due to the difficulties associated with the step-by-step addition of individual monomers to achieve high molecular weights. Here we report the copolymers of α-hydroxy acids, poly(phenyllactic-co-lactic acid) (PcL) built via the cross-convergent method from four dyads of monomers as constituent units. Our proposed method allows scalable synthesis of sequence-defined PcL in a minimal number of coupling steps from reagents in stoichiometric amounts. Digital information can be stored in an aperiodic sequence of PcL, which can be fully retrieved as binary code by mass spectrometry sequencing. The information storage density (bit/Da) of PcL is 50% higher than DNA, and the storage capacity of PcL can also be increased by adjusting the molecular weight (~38 kDa).

Sign in / Sign up

Export Citation Format

Share Document