Input-to-state stability analysis for memristive BAM neural networks with variable time delays

In this paper, we are concerned with input-to-state stability of a class of memristive bidirectional associative memory (BAM) neural networks with variable time delays. Based on a nonsmooth analysis and set-valued maps, some novel sufficient conditions are obtained for the input-to-state stability of such networks, which extended some known results as particular cases. Finally, a numerical example is presented to illustrate the feasibility and effectiveness of our results.     

忆阻突触耦合环形Hopfield神经网络动力学分析及其电路实现

提出一种新型忆阻器模型, 利用标准非线性理论分析三个忆阻特性, 并设计模拟电路。基于忆阻突触, 构建一个忆阻突触耦合环形Hopfield神经网络模型。采用分岔图、李雅普诺夫指数谱、时序图等方法, 揭示与忆阻突触密切相关的特殊动力学行为。数值仿真表明: 在忆阻突触权重的影响下, 它能够产生多种对称簇发放电模式和复杂的混沌行为。实现了该忆阻环形神经网络的模拟等效电路, 并由PSIM电路仿真验证MATLAB数值仿真的正确性。     

SPICE modeling of flux-controlled unipolar memristive devices

Unipolar memristive devices are an important kind of resistive switching devices. However, few circuit models of them have been proposed. In this paper, we propose the SPICE modeling of flux-controlled unipolar memristive devices based on the memristance versus state map. Using our model, the flux thresholds, ON and OFF resistance, and compliance current can easily be set as model parameters. We simulate the model in HSPICE using model parameters abstracted from real devices, and the simulation results show that the proposed model caters to the real device data very well, thus demonstrating that the model is correct. Using the same modeling methodology, the SPICE model of charge-controlled unipolar memristive devices could also be developed. The proposed model could be used to model resistive memory cells, logical gates as well as synapses in artificial neural networks.     

忆阻Hopfield神经网络动力学分析及其电路实现

提出一种超多稳态忆阻Hopfield神经网络, 它仅包含3个神经元和一个多稳态忆阻突触。从理论上分析神经网络的耗散性和平衡点的稳定性, 并利用分岔图、李雅普诺夫指数谱和相位图等数值方法分析不同忆阻突触耦合强度对神经网络动力学的影响。网络参数固定时, 揭示与初始状态值密切相关的超多稳态性动力学行为。最后, 设计忆阻Hopfield神经网络的模拟等效电路, 并通过PSIM电路仿真验证MATLAB数值仿真结果。     

Memory feedback finite-time control for memristive neutral-type neural networks with quantization

The drive-response of memory feedback control design for memristor neural networks of neutral type over finite-time domain is scrutinized in this paper. Notably, the main purpose of this work is to synthesize memory feedback controller in the presence of logarithmic quantizer and actuator saturation to guarantee the finite-time boundedness of the resulting memristive neural networks. On basis of proper Lyapunov–Krasovskii-functional and linear matrix inequalities, new sufficient criterian is established to assure the delay-dependent finite-time stabilization criteria for the addressed network model. Also, by solving the developed linear matrix inequalities, the finite-time memory feedback control law gain matrices could be attained. Eventually, the validations of the proposed mechanism are ultimately explored through two numerical examples.     

Optically-controlled resistive switching effectsof CdS nanowire memtransistor

Since it was proposed, memtransistors have been a leading candidate with powerful capabilities in the field of neural morphological networks. A memtransistor is an emerging structure combining the concepts of a memristor and a field-effect transistor with low-dimensional materials, so that both optical excitation and electrical stimuli can be used to modulate the memristive characteristics, which make it a promising multi-terminal hybrid device for synaptic structures. In this paper, a single CdS nanowire memtransistor has been constructed by the micromechanical exfoliation and alignment lithography methods. It is found that the CdS memtransistor has good non-volatile bipolar memristive characteristics, and the corresponding switching ratio is as high as 10⁶ in the dark. While under illumination, the behavior of the CdS memtransistor is similar to that of a transistor or a memristor depending on the incident wavelengths, and the memristive switching ratio varies in the range of 10 to 10⁵ with the increase of the incident wavelength in the visible light range. In addition, the optical power is also found to affect the memristive characteristics of the device. All of these can be attributed to the modulation of the potential barrier by abundant surface states of nanowires and the illumination influences on the carrier concentrations in nanowires.     

忆阻器及其阻变机理研究进展

忆阻器是除电阻、电容、电感之外的第四种电路元件,在信息存储、逻辑运算和神经网络等研究领域具有重要的应用前景.本文综述了忆阻器以及忆阻器材料的研究进展,主要介绍了忆阻器的内涵与特征、阻变机理、材料类型以及应用前景,指出了目前忆阻器研究中需要关注的主要问题,并对以后的发展趋势进行了展望.     

基于串并联磁控忆阻器的耦合行为研究

忆阻器是纳米级器件, 其功耗低, 集成度高, 有着巨大的应用潜能. 单个器件具有丰富的电学性质, 其串并联电路更展现了丰富的动力学行为. 然而, 忆阻器在高密度集成的环境下, 其耦合效应不可忽视. 因此, 本文首先基于磁控忆阻器推导了耦合忆阻器的数学模型. 其次, 在考虑不同极性连接和耦合强度的前提下, 讨论两个磁控忆阻器串并联的耦合情况, 进行了详细的理论分析, 并通过数值仿真探索了耦合效应对忆阻系统的影响. 同时, 设计了基于Matlab的图形用户界面, 直观地展示了不同参数下的耦合特性曲线. 进一步, 本文展示了有无耦合情况下, 初始阻值对忆阻器正常工作范围的影响. 最后, 构建耦合忆阻器的Pspice仿真器, 从电路的角度再次验证了忆阻器间的耦合效应. 实验结果表明: 同极性耦合增强了阻值的改变, 相反极性的耦合减缓了阻值的改变. 这些动力学特性可以很好地应用于忆阻网络中, 也为全面考虑忆阻系统电路的设计提供了强大的理论基础.     

Pulse coding off-chip learning algorithm for memristive artificial neural network

Memristive neural network has attracted tremendous attention since the memristor array can perform parallel multiply-accumulate calculation (MAC) operations and memory-computation operations as compared with digital CMOS hardware systems. However, owing to the variability of the memristor, the implementation of high-precision neural network in memristive computation units is still difficult. Existing learning algorithms for memristive artificial neural network (ANN) is unable to achieve the performance comparable to high-precision by using CMOS-based system. Here, we propose an algorithm based on off-chip learning for memristive ANN in low precision. Training the ANN in the high-precision in digital CPUs and then quantifying the weight of the network to low precision, the quantified weights are mapped to the memristor arrays based on VTEAM model through using the pulse coding weight-mapping rule. In this work, we execute the inference of trained 5-layers convolution neural network on the memristor arrays and achieve an accuracy close to the inference in the case of high precision (64-bit). Compared with other algorithms-based off-chip learning, the algorithm proposed in the present study can easily implement the mapping process and less influence of the device variability. Our result provides an effective approach to implementing the ANN on the memristive hardware platform.     

场耦合忆阻神经网络的电活动

神经元之间除了突触耦合,还存在磁通耦合.因此在传统的神经元模型中引入磁通量,并研究场耦合下神经网络的放电活动具有实际意义.建立一个含场耦合的Hodgkin-Huxley忆阻神经网络,引入神经元节点之间的距离权重,用磁通量描述时变电磁场,采用磁控忆阻器实现膜电位和磁通量之间的耦合.探讨距离权重和系统大小对神经网络放电活动的影响.研究发现,随着权重增大,神经网络放电活动增强,且系统规模越大,诱导神经元兴奋性的权重阈值越大,系统大小不影响神经网络活性随距离权重变化的规律.在不同的权重值下,神经网络活性随系统大小变化的规律明显不同.研究表明,距离权重和系统大小对含场耦合的忆阻神经网络放电活动有重要影响,其中距离权重起主导作用.     

Mean Field Theory of Self-Organizing Memristive Connectomes

Biological neuronal networks are characterized by nonlinear interactions and complex connectivity. Given the growing impetus to build neuromorphic computers, understanding physical devices that exhibit structures and functionalities similar to biological neural networks is an important step toward this goal. Self-organizing circuits of nanodevices are at the forefront of the research in neuromorphic computing, as their behavior mimics synaptic plasticity features of biological neuronal circuits. However, an effective theory to describe their behavior is lacking. This study provides for the first time an effective mean field theory for the emergent voltage-induced polymorphism of circuits of a nanowire connectome, showing that the behavior of these circuits can be explained by a low-dimensional dynamical equation. The equation can be derived from the microscopic dynamics of a single memristive junction in analytical form. The effective model is tested on experiments of nanowire networks and show that it fits both the potentiation and depression of these synapse-mimicking circuits. It is shown that this theory applies beyond the case of nanowire networks by formulating a general mean-field theory of conductance transitions in self-organizing memristive connectomes.     

Synchronization in an array of linearly stochastically coupled networks with time delays

In this paper, the complete synchronization problem is investigated in an array of linearly stochastically coupled identical networks with time delays. The stochastic coupling term, which can reflect a more realistic dynamical behavior of coupled systems in practice, is introduced to model a coupled system, and the influence from the stochastic noises on the array of coupled delayed neural networks is studied thoroughly. Based on a simple adaptive feedback control scheme and some stochastic analysis techniques, several sufficient conditions are developed to guarantee the synchronization in an array of linearly stochastically coupled neural networks with time delays. Finally, an illustrate example with numerical simulations is exploited to show the effectiveness of the theoretical results.     

An improved memristor model for brain-inspired computing

Memristors, as memristive devices, have received a great deal of interest since being fabricated by HP labs. The forgetting effect that has significant influences on memristors' performance has to be taken into account when they are employed. It is significant to build a good model that can express the forgetting effect well for application researches due to its promising prospects in brain-inspired computing. Some models are proposed to represent the forgetting effect but do not work well. In this paper, we present a novel window function, which has good performance in a drift model. We analyze the deficiencies of the previous drift diffusion models for the forgetting effect and propose an improved model. Moreover,the improved model is exploited as a synapse model in spiking neural networks to recognize digit images. Simulation results show that the improved model overcomes the defects of the previous models and can be used as a synapse model in brain-inspired computing due to its synaptic characteristics. The results also indicate that the improved model can express the forgetting effect better when it is employed in spiking neural networks, which means that more appropriate evaluations can be obtained in applications.     

Robust adaptive synchronization of chaotic neural networks by slide technique

In this paper, we focus on the robust adaptive synchronization between two coupled chaotic neural networks with all the parameters unknown and time-varying delay. In order to increase the robustness of the two coupled neural networks, the key idea is that a sliding-mode-type controller is employed. Moreover, without the estimate values of the network unknown parameters taken as an updating object, a new updating object is introduced in the constructing of controller. Using the proposed controller, without any requirements for the boundedness, monotonicity and differentiability of activation functions, and symmetry of connections, the two coupled chaotic neural networks can achieve global robust synchronization no matter what their initial states are. Finally, the numerical simulation validates the effectiveness and feasibility of the proposed technique.     

Simulation of synaptic coupling of neuron-like generators via a memristive device

A physical model of synaptically coupled neuron-like generators interacting via a memristive device has been presented. The model simulates the synaptic transmission of pulsed signals between brain neurons. The action on the receiving generator has been performed via a memristive device that demonstrates adaptive behavior. It has been established that the proposed coupling channel provides the forced synchronization with the parameters depending on the memristive device sensitivity. Synchronization modes 1: 1 and 2: 1 have been experimentally observed.     

Adaptive synchronization in an array of asymmetric coupled neural networks

This paper investigates the global synchronization in an array of linearly coupled neural networks with constant and delayed coupling. By a simple combination of adaptive control and linear feedback with the updated laws, some sufficient conditions are derived for global synchronization of the coupled neural networks. The coupling configuration matrix is assumed to be asymmetric, which is more coincident with the realistic network. It is shown that the approaches developed here extend and improve the earlier works. Finally, numerical simulations are presented to demonstrate the effectiveness of the theoretical results.     

High-performance artificial neurons based on Ag/MXene/GST/Pt threshold switching memristors

Threshold switching (TS) memristors can be used as artificial neurons in neuromorphic systems due to their continuous conductance modulation, scalable and energy-efficient properties. In this paper, we propose a low power artificial neuron based on the Ag/MXene/GST/Pt device with excellent TS characteristics, including a low set voltage (0.38 V) and current (200 nA), an extremely steep slope (< 0.1 mV/dec), and a relatively large off/on ratio (> 10³). Besides, the characteristics of integrate and fire neurons that are indispensable for spiking neural networks have been experimentally demonstrated. Finally, its memristive mechanism is interpreted through the first-principles calculation depending on the electrochemical metallization effect.     

Finite-time Mittag—Leffler synchronization of fractional-order complex-valued memristive neural networks with time delay

Without dividing the complex-valued systems into two real-valued ones, a class of fractional-order complex-valued memristive neural networks (FCVMNNs) with time delay is investigated. Firstly, based on the complex-valued sign function, a novel complex-valued feedback controller is devised to research such systems. Under the framework of Filippov solution, differential inclusion theory and Lyapunov stability theorem, the finite-time Mittag—Leffler synchronization (FTMLS) of FCVMNNs with time delay can be realized. Meanwhile, the upper bound of the synchronization settling time (SST) is less conservative than previous results. In addition, by adjusting controller parameters, the global asymptotic synchronization of FCVMNNs with time delay can also be realized, which improves and enrich some existing results. Lastly, some simulation examples are designed to verify the validity of conclusions.     

Synchronization of stochastically hybrid coupled neural networks with coupling discrete and distributed time-varying delays

A general model of linearly stochastically coupled identical connected neural networks with hybrid coupling is proposed, which is composed of constant coupling, coupling discrete time-varying delay and coupling distributed timevarying delay. All the coupling terms are subjected to stochastic disturbances described in terms of Brownian motion, which reflects a more realistic dynamical behaviour of coupled systems in practice. Based on a simple adaptive feedback controller and stochastic stability theory, several sufficient criteria are presented to ensure the synchronization of linearly stochastically coupled complex networks with coupling mixed time-varying delays. Finally, numerical simulations illustrated by scale-free complex networks verify the effectiveness of the proposed controllers.     

A spintronic memristive circuit on the optimized RBF-MLP neural network

A radial basis function network (RBF) has excellent generalization ability and approximation accuracy when its parameters are set appropriately. However, when relying only on traditional methods, it is difficult to obtain optimal network parameters and construct a stable model as well. In view of this, a novel radial basis neural network (RBF-MLP) is proposed in this article. By connecting two networks to work cooperatively, the RBF's parameters can be adjusted adaptively by the structure of the multi-layer perceptron (MLP) to realize the effect of the backpropagation updating error. Furthermore, a genetic algorithm is used to optimize the network's hidden layer to confirm the optimal neurons (basis function) number automatically. In addition, a memristive circuit model is proposed to realize the neural network's operation based on the characteristics of spin memristors. It is verified that the network can adaptively construct a network model with outstanding robustness and can stably achieve 98.33% accuracy in the processing of the Modified National Institute of Standards and Technology (MNIST) dataset classification task. The experimental results show that the method has considerable application value.