scholarly journals Mapping the BCPNN Learning Rule to a Memristor Model

2021 ◽  
Vol 15 ◽  
Author(s):  
Deyu Wang ◽  
Jiawei Xu ◽  
Dimitrios Stathis ◽  
Lianhao Zhang ◽  
Feng Li ◽  
...  
Keyword(s):  
Neural Network ◽  
Correlation Coefficient ◽  
Analog Circuit ◽  
Memory Function ◽  
Learning Rule ◽  
Signal Design ◽  
State Variables ◽  
Von Neumann ◽  
Bayesian Confidence Propagation ◽  
And Storage

The Bayesian Confidence Propagation Neural Network (BCPNN) has been implemented in a way that allows mapping to neural and synaptic processes in the human cortexandhas been used extensively in detailed spiking models of cortical associative memory function and recently also for machine learning applications. In conventional digital implementations of BCPNN, the von Neumann bottleneck is a major challenge with synaptic storage and access to it as the dominant cost. The memristor is a non-volatile device ideal for artificial synapses that fuses computation and storage and thus fundamentally overcomes the von Neumann bottleneck. While the implementation of other neural networks like Spiking Neural Network (SNN) and even Convolutional Neural Network (CNN) on memristor has been studied, the implementation of BCPNN has not. In this paper, the BCPNN learning rule is mapped to a memristor model and implemented with a memristor-based architecture. The implementation of the BCPNN learning rule is a mixed-signal design with the main computation and storage happening in the analog domain. In particular, the nonlinear dopant drift phenomenon of the memristor is exploited to simulate the exponential decay of the synaptic state variables in the BCPNN learning rule. The consistency between the memristor-based solution and the BCPNN learning rule is simulated and verified in Matlab, with a correlation coefficient as high as 0.99. The analog circuit is designed and implemented in the SPICE simulation environment, demonstrating a good emulation effect for the BCPNN learning rule with a correlation coefficient as high as 0.98. This work focuses on demonstrating the feasibility of mapping the BCPNN learning rule to in-circuit computation in memristor. The feasibility of the memristor-based implementation is evaluated and validated in the paper, to pave the way for a more efficient BCPNN implementation, toward a real-time brain emulation engine.

Determination of thermodynamic state variables of liquids from its microscopic structure using an artificial neural network

Soft Matter ◽  
2020 ◽  
Author(s):  
Ulices Que-Salinas ◽  
Pedro Ezequiel Ramirez-Gonzalez ◽  
Alexis Torres-Carbajal
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Artificial Neural Network ◽  
Distribution Function ◽  
Microscopic Structure ◽  
Machine Learning Method ◽  
Learning Method ◽  
State Variables ◽  
Thermodynamic State

In this work we implement a machine learning method to predict the thermodynamic state of a liquid using only its microscopic structure provided by the radial distribution function (RDF). The...

scholarly journals A Neural Network Decomposition Algorithm for Mapping on Crossbar-Based Computing Systems

Electronics ◽  
2020 ◽  
Vol 9 (9) ◽  
pp. 1526 ◽  
Author(s):  
Choongmin Kim ◽  
Jacob A. Abraham ◽  
Woochul Kang ◽  
Jaeyong Chung
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Predictive Performance ◽  
Fixed Number ◽  
Decomposition Algorithm ◽  
Computing Systems ◽  
Network Decomposition ◽  
Von Neumann ◽  
Analog Computing ◽  
The Neural Network

Crossbar-based neuromorphic computing to accelerate neural networks is a popular alternative to conventional von Neumann computing systems. It is also referred as processing-in-memory and in-situ analog computing. The crossbars have a fixed number of synapses per neuron and it is necessary to decompose neurons to map networks onto the crossbars. This paper proposes the k-spare decomposition algorithm that can trade off the predictive performance against the neuron usage during the mapping. The proposed algorithm performs a two-level hierarchical decomposition. In the first global decomposition, it decomposes the neural network such that each crossbar has k spare neurons. These neurons are used to improve the accuracy of the partially mapped network in the subsequent local decomposition. Our experimental results using modern convolutional neural networks show that the proposed method can improve the accuracy substantially within about 10% extra neurons.

scholarly journals Prediction of SYM-H index during large storms by NARX neural network from IMF and solar wind data

2010 ◽  
Vol 28 (2) ◽  
pp. 381-393 ◽  
Author(s):  
L. Cai ◽  
S. Y. Ma ◽  
Y. L. Zhou
Keyword(s):  
Neural Network ◽  
Solar Wind ◽  
Correlation Coefficient ◽  
Characteristic Time ◽  
Ring Current ◽  
Wind Data ◽  
Distinct Advantage ◽  
Elman Network ◽  
H Index ◽  
Narx Neural Network

Abstract. Similar to the Dst index, the SYM-H index may also serve as an indicator of magnetic storm intensity, but having distinct advantage of higher time-resolution. In this study the NARX neural network has been used for the first time to predict SYM-H index from solar wind (SW) and IMF parameters. In total 73 time intervals of great storm events with IMF/SW data available from ACE satellite during 1998 to 2006 are used to establish the ANN model. Out of them, 67 are used to train the network and the other 6 samples for test. Additionally, the NARX prediction model is also validated using IMF/SW data from WIND satellite for 7 great storms during 1995–1997 and 2005, as well as for the July 2000 Bastille day storm and November 2001 superstorm using Geotail and OMNI data at 1 AU, respectively. Five interplanetary parameters of IMF Bz, By and total B components along with proton density and velocity of solar wind are used as the original external inputs of the neural network to predict the SYM-H index about one hour ahead. For the 6 test storms registered by ACE including two super-storms of min. SYM-H<−200 nT, the correlation coefficient between observed and NARX network predicted SYM-H is 0.95 as a whole, even as high as 0.95 and 0.98 with average relative variance of 13.2% and 7.4%, respectively, for the two super-storms. The prediction for the 7 storms with WIND data is also satisfactory, showing averaged correlation coefficient about 0.91 and RMSE of 14.2 nT. The newly developed NARX model shows much better capability than Elman network for SYM-H prediction, which can partly be attributed to a key feedback to the input layer from the output neuron with a suitable length (about 120 min). This feedback means that nearly real information of the ring current status is effectively directed to take part in the prediction of SYM-H index by ANN. The proper history length of the output-feedback may mainly reflect on average the characteristic time of ring current decay which involves various decay mechanisms with ion lifetimes from tens of minutes to tens of hours. The Elman network makes feedback from hidden layer to input only one step, which is of 5 min for SYM-H index in this work and thus insufficient to catch the characteristic time length.

scholarly journals The BP Network Classification Leafminer-Infected Leaves Based on the Fractal Dimension

2011 ◽  
Vol 1 ◽  
pp. 163-167
Author(s):  
Da Ke Wu ◽  
Chun Yan Xie
Keyword(s):  
Neural Network ◽  
Fractal Dimension ◽  
Correlation Coefficient ◽  
Bp Neural Network ◽  
Input Data ◽  
Bp Network ◽  
Neural Network Prediction ◽  
Infected Tomato ◽  
Network Prediction ◽  
Hidden Nodes

Leafminer is one of pest of many vegetables, and the damage may cover so much of the leaf that the plant is unable to function, and yields are noticeably decreased. In order to get the information of the pest in the vegetable before the damage was not serious, this research used a BP neural network to classify the leafminer-infected tomato leaves, and the fractal dimension of the leaves was the input data of the BP neural network. Prediction results showed that when the number of FD was 21 and the hidden nodes of BP neural network were 21, the detection performance of the model was good and the correlation coefficient (r) was 0.836. Thus, it is concluded that the FD is an available technique for the detection of disease level of leafminer on tomato leaves.

KNT-artificial neural network model for flux prediction of ultrafiltration membrane producing drinking water

2004 ◽  
Vol 50 (8) ◽  
pp. 103-110 ◽  
Author(s):  
H.K. Oh ◽  
M.J. Yu ◽  
E.M. Gwon ◽  
J.Y. Koo ◽  
S.G. Kim ◽  
...  
Keyword(s):  
Neural Network ◽  
Experimental Data ◽  
Artificial Neural Network ◽  
Correlation Coefficient ◽  
Network Model ◽  
Operating Conditions ◽  
Transmembrane Pressure ◽  
Feed Water ◽  
Permeate Flux ◽  
Artificial Neural

This paper describes the prediction of flux behavior in an ultrafiltration (UF) membrane system using a Kalman neuro training (KNT) network model. The experimental data was obtained from operating a pilot plant of hollow fiber UF membrane with groundwater for 7 months. The network was trained using operating conditions such as inlet pressure, filtration duration, and feed water quality parameters including turbidity, temperature and UV254. Pre-processing of raw data allowed the normalized input data to be used in sigmoid activation functions. A neural network architecture was structured by modifying the number of hidden layers, neurons and learning iterations. The structure of KNT-neural network with 3 layers and 5 neurons allowed a good prediction of permeate flux by 0.997 of correlation coefficient during the learning phase. Also the validity of the designed model was evaluated with other experimental data not used during the training phase and nonlinear flux behavior was accurately estimated with 0.999 of correlation coefficient and a lower error of prediction in the testing phase. This good flux prediction can provide preliminary criteria in membrane design and set up the proper cleaning cycle in membrane operation. The KNT-artificial neural network is also expected to predict the variation of transmembrane pressure during filtration cycles and can be applied to automation and control of full scale treatment plants.

Pattern Storage, Bifurcations, and Groupwise Correlation Structure of an Exactly Solvable Asymmetric Neural Network Model

2018 ◽  
Vol 30 (5) ◽  
pp. 1258-1295 ◽  
Author(s):  
Diego Fasoli ◽  
Anna Cattani ◽  
Stefano Panzeri
Keyword(s):  
Neural Network ◽  
Evolution Equations ◽  
Learning Rule ◽  
Network Size ◽  
Oscillatory Behavior ◽  
Spin Model ◽  
Membrane Potentials ◽  
Neural Network Models ◽  
Conditional Probability Distribution ◽  
Firing Rates

Despite their biological plausibility, neural network models with asymmetric weights are rarely solved analytically, and closed-form solutions are available only in some limiting cases or in some mean-field approximations. We found exact analytical solutions of an asymmetric spin model of neural networks with arbitrary size without resorting to any approximation, and we comprehensively studied its dynamical and statistical properties. The network had discrete time evolution equations and binary firing rates, and it could be driven by noise with any distribution. We found analytical expressions of the conditional and stationary joint probability distributions of the membrane potentials and the firing rates. By manipulating the conditional probability distribution of the firing rates, we extend to stochastic networks the associating learning rule previously introduced by Personnaz and coworkers. The new learning rule allowed the safe storage, under the presence of noise, of point and cyclic attractors, with useful implications for content-addressable memories. Furthermore, we studied the bifurcation structure of the network dynamics in the zero-noise limit. We analytically derived examples of the codimension 1 and codimension 2 bifurcation diagrams of the network, which describe how the neuronal dynamics changes with the external stimuli. This showed that the network may undergo transitions among multistable regimes, oscillatory behavior elicited by asymmetric synaptic connections, and various forms of spontaneous symmetry breaking. We also calculated analytically groupwise correlations of neural activity in the network in the stationary regime. This revealed neuronal regimes where, statistically, the membrane potentials and the firing rates are either synchronous or asynchronous. Our results are valid for networks with any number of neurons, although our equations can be realistically solved only for small networks. For completeness, we also derived the network equations in the thermodynamic limit of infinite network size and we analytically studied their local bifurcations. All the analytical results were extensively validated by numerical simulations.

scholarly journals Prediction of Wave Power Generation Using a Convolutional Neural Network with Multiple Inputs

Energies ◽  
2018 ◽  
Vol 11 (8) ◽  
pp. 2097 ◽  
Author(s):  
Chenhua Ni ◽  
Xiandong Ma
Keyword(s):  
Neural Network ◽  
Power Generation ◽  
Convolutional Neural Network ◽  
Wave Height ◽  
Regression Models ◽  
Cost Effective ◽  
Wave Power ◽  
Marine Energy ◽  
Artificial Network ◽  
And Storage

Successful development of a marine wave energy converter (WEC) relies strongly on the development of the power generation device, which needs to be efficient and cost-effective. An innovative multi-input approach based on the Convolutional Neural Network (CNN) is investigated to predict the power generation of a WEC system using a double-buoy oscillating body device (OBD). The results from the experimental data show that the proposed multi-input CNN performs much better at predicting results compared with the conventional artificial network and regression models. Through the power generation analysis of this double-buoy OBD, it shows that the power output has a positive correlation with the wave height when it is higher than 0.2 m, which becomes even stronger if the wave height is higher than 0.6 m. Furthermore, the proposed approach associated with the CNN algorithm in this study can potentially detect the changes that could be due to presence of anomalies and therefore be used for condition monitoring and fault diagnosis of marine energy converters. The results are also able to facilitate controlling of the electricity balance among energy conversion, wave power produced and storage.

Sign in / Sign up

Export Citation Format

Share Document