Memristive Synapses Research Articles

Low-power artificial neuron networks with enhanced synaptic functionality using dual transistor and dual memristor.

Artificial neurons with bio-inspired firing patterns have the potential to significantly improve the performance of neural network computing. The most significant component of an artificial neuron circuit is a large amount of energy consumption. Recent literature has proposed memristors as a promising option for synaptic implementation. In contrast, implementing memristive circuitry through neuron hardware presents significant challenges and is a relevant research topic. This paper describes an efficient circuit-level mixed CMOS memristor artificial neuron network with a memristor synapse model. From this perspective, the paper describes the design of artificial neurons in standard CMOS technology with low power utilization. The neuron circuit response is a modified version of the Morris-Lecar theoretical model. The suggested circuit employs memristor-based artificial neurons with Dual Transistor and Dual Memristor (DTDM) synapse circuit. The proposed neuron network produces a high spiking frequency and low power consumption. According to our research, a memristor-based Morris Lecar (ML) neuron with a DTDM synapse circuit consumes 12.55 pW of power, the spiking frequency is 22.72 kHz, and 2.13 fJ of energy per spike. The simulations were carried out using the Spectre tool with 45 nm CMOS technology.

High-Performance Memristive Synapse Based on Space-Charge-Limited Conduction in LiNbO3.

Advancing neuromorphic computing technology requires the development of versatile synaptic devices. In this study, we fabricated a high-performance Al/LiNbO3/Pt memristive synapse and emulated various synaptic functions using its primary key operating mechanism, known as oxygen vacancy-mediated valence charge migration (VO-VCM). The voltage-controlled VO-VCM induced space-charge-limited conduction and self-rectifying asymmetric hysteresis behaviors. Moreover, the device exhibited voltage pulse-tunable multi-state memory characteristics because the degree of VO-VCM was dependent on the applied pulse parameters (e.g., polarity, amplitude, width, and interval). As a result, synaptic functions such as short-term memory, dynamic range-tunable long-term memory, and spike time-dependent synaptic plasticity were successfully demonstrated by modulating those pulse parameters. Additionally, simulation studies on hand-written image pattern recognition confirmed that the present device performed with high accuracy, reaching up to 95.2%. The findings suggest that the VO-VCM-based Al/LiNbO3/Pt memristive synapse holds significant promise as a brain-inspired neuromorphic device.

Collective behavior of an adapting synapse-based neuronal network with memristive effect and randomness.

This study delves into the examination of a network of adaptive synapse neurons characterized by a small-world network topology connected through electromagnetic flux and infused with randomness. First, this research extensively explores the existence of the global multi-stability of a single adaptive synapse-based neuron model with magnetic flux. The non-autonomous neuron model exhibits periodically switchable equilibrium states that are strongly related to the transitions between stable and unstable points in every whole periodic cycle, leading to the creation of global multi-stability. Various numerical measures, including bifurcation plots, phase plots, and basin of attraction, illustrate the intricate dynamics of diverse coexisting global firing activities. Moreover, the model is extended by coupling two neurons with a memristive synapse. The dynamics of the coupled neurons model are showcased with the help of largest Lyapunov exponents, and synchronized dynamics are viewed with the help of mean average error. Next, we consider a regular network of neurons connected to their nearest neighbors through the memristive synapse. We then reconstruct it into a small-world network by increasing the randomness in the rewiring links. Consequently, we observed collective behavior influenced by the number of neighborhood connections, coupling strength, and rewiring probability. We used spatio-temporal patterns, recurrence plots, as well as global-order parameters to verify the reported results.

Dynamic Analysis and FPGA Implementation of Fractional-Order Hopfield Networks with Memristive Synapse

Memristors have become important components in artificial synapses due to their ability to emulate the information transmission and memory functions of biological synapses. Unlike their biological counterparts, which adjust synaptic weights, memristor-based artificial synapses operate by altering conductance or resistance, making them useful for enhancing the processing capacity and storage capabilities of neural networks. When integrated into systems like Hopfield neural networks, memristors enable the study of complex dynamic behaviors, such as chaos and multistability. Moreover, fractional calculus is significant for their ability to model memory effects, enabling more accurate simulations of complex systems. Fractional-order Hopfield networks, in particular, exhibit chaotic and multistable behaviors not found in integer-order models. By combining memristors with fractional-order Hopfield neural networks, these systems offer the possibility of investigating different dynamic phenomena in artificial neural networks. This study investigates the dynamical behavior of a fractional-order Hopfield neural network (HNN) incorporating a memristor with a piecewise segment function in one of its synapses, highlighting the impact of fractional-order derivatives and memristive synapses on the stability, robustness, and dynamic complexity of the system. Using a network of four neurons as a case study, it is demonstrated that the memristive fractional-order HNN exhibits multistability, coexisting chaotic attractors, and coexisting limit cycles. Through spectral entropy analysis, the regions in the initial condition space that display varying degrees of complexity are mapped, highlighting those areas where the chaotic series approach a pseudo-random sequence of numbers. Finally, the proposed fractional-order memristive HNN is implemented on a Field-Programmable Gate Array (FPGA), demonstrating the feasibility of real-time hardware realization.

Influences of short-term and long-term plasticity of memristive synapse on firing activity of neuronal network

Abstract Synaptic plasticity can greatly affect the firing behavior of neural networks, and it specifically refers to changes in the strength, morphology, and function of synaptic connections. In this paper, a novel memristor model, which can be configured as a volatile and nonvolatile memristor by adjusting its internal parameter, is proposed to mimic the short-term and long-term synaptic plasticity. Then, a bi-neuron network model, with the proposed memristor serving as a coupling synapse and the external electromagnetic radiation being emulated by the flux-controlled memristors, is established to elucidate the effects of short-term and long-term synaptic plasticity on firing activity of the neuron network. The resultant seven-dimensional (7D) neuron network has no equilibrium point and its hidden dynamical behavior is revealed by phase diagram, time series, bifurcation diagram, Lyapunov exponent spectrum, and two-dimensional (2D) dynamic map. Our results show the short-term and long-term plasticity can induce different bifurcation scenarios when the coupling strength increases. In addition, memristor synaptic plasticity has a great influence on the distribution of firing patterns in the parameter space. More interestingly, when exploring the synchronous firing behavior of two neurons, the two neurons can gradually achieve phase synchronization as the coupling strength increases along the opposite directions under two different memory attributes. Finally, a microcontroller-based hardware system is implemented to verify the numerical simulation results.

Transition and bifurcation mechanism of firing activities in memristor synapse-coupled Hindmarsh–Rose bi-neuron model

Memristor, especially the locally active one, is widely used in emulating the synapse-inspired activities in neural networks, in which the connection between memristor properties and neuronal electrical activities requires further investigation. In this paper, a bi-neuron model is constructed by bi-directionally connecting a locally active memristor between the membrane potentials of two deterministic 3D HR neuron models. Complex spiking/bursting firing activities and their transitions are disclosed under different memristor control parameters. The bifurcation mechanisms of the revealed synchronous and asynchronous firing activities are explored inside each individual neuron by taking the externally input state variables as modulation parameters. These results demonstrate the crucial effect of the locally active memristor synapse on firing activities of its coupled network. FPGA-based digital circuit experiment is finally performed to verify the numerical results. This research is beneficial to understanding, control and application of firing activities in neural networks.

A Self-Driven Ga2O3 Memristor Synapse for Humanoid Robot Learning.

In recent years, the rapid development of brain-inspired neuromorphic systems has created an imperative demand for artificial photonic synapses that operate with low power consumption. In this study, a self-driven memristor synapse based on gallium oxide (Ga2O3) nanowires is proposed and demonstrated successfully. This memristor synapse is capable of emulating a range of functionalities of biological synapses when exposed to 255nm light stimulation. These functionalities encompass peak time-dependent plasticity, pulse facilitation, and memory learning capabilities. It exhibits an ultrahigh paired-pulse facilitation index of 158, indicating exceptional learning performance. The transition from short-term memory to long-term memory can be attributed to the remarkable relearning capabilities. Furthermore, the potential applications of the memristor synapse is showcased through the successful manipulation of a humanoid intelligent robot. Upon establishing artificial intelligence (AI) systems, the control commands originating from the synaptic device can drive the humanoid robot to perform various actions. Based on the memristor synapses, the autonomous feedback system of the humanoid robot facilitates a good collaboration between robotic actions and bio-inspired light perception. Therefore, this research opens up an effective way to advance the development of neuromorphic computing technologies, AI systems, and intelligent robots that demand ultra-low energy consumption.

Frequency chimera state induced by time delays in FitzHugh-Nagumo neural networks

The delay of information transmission is an inherent factor of the nervous systems, which has great influences on their collective dynamic behaviors. In this paper, we constructed a ring neural network using FitzHugh-Nagumo (FHN) neuron model as the network node and memristive synapses as the connection mode. Our primary focus was on investigating the effects of time delay and coupling strength on the firing frequency of neurons. Simulation results revealed that the frequency chimera state could be induced in the neural network with appropriate time delay, which is a new type of chimera state characterized by firing frequency rather than traditional membrane potential. By adjusting the time delay properly, the neural network can also display multi-cluster frequency chimera states that coexisted with various incoherent regions and coherent regions. Meanwhile, we exhibit that initial value and coupling strength could have great influences on the effects of time delay on inducing frequency chimera state of the nervous systems.

Deep brain stimulation and lag synchronization in a memristive two-neuron network

In the pursuit of potential treatments for neurological disorders and the alleviation of patient suffering, deep brain stimulation (DBS) has been utilized to intervene or investigate pathological neural activities. To explore the exact mechanism of how DBS works, a memristive two-neuron network considering DBS is newly proposed in this work. This network is implemented by coupling two-dimensional Morris–Lecar neuron models and using a memristor synaptic synapse to mimic synaptic plasticity. The complex bursting activities and dynamical effects are revealed numerically through dynamical analysis. By examining the synchronous behavior, the desynchronization mechanism of the memristor synapse is uncovered. The study demonstrates that synaptic connections lead to the appearance of time-lagged or asynchrony in completely synchronized firing activities. Additionally, the memristive two-neuron network is implemented in hardware based on FPGA, and experimental results confirm the abundant neuronal electrical activities and chaotic dynamical behaviors. This work offers insights into the potential mechanisms of DBS intervention in neural networks.

Dynamics of a two-neuron hopfield neural network: Memristive synapse and autapses and impact of fractional order

There are numerous studies on Hopfield neural networks with electromagnetic induction using memristors in either autaptic or synaptic connections. In this study, we explore a novel scenario where all connections are influenced by electromagnetic induction. We investigate and compare the network’s dynamics with one memristive autapse, two memristive autapses, and a memristive synapse. The results indicate that having two memristive autapses instead of one increases the dynamical range, leading to chaotic dynamics in unequal autaptic strengths. In contrast, in the presence of the memristive synapse, chaos can emerge only in very strong synaptic strength. Using fractional-order derivatives can transform the periodic attractor of the integer-order model into a chaotic one in some parameters. Furthermore, incorporating more memristors leads to chaos at lower fractional orders.

Parametric controllable planar multi-scroll chaotic attractors in a 3-D memristive tabu learning single neuron model

In this paper, a three-dimensional (3D) autonomous tabu learning single neuron model is proposed, which is achieved by using a sinusoidal activation function and introducing a memristor synapse. This model exhibits the remarkable capability to produce a series of planar multi-scroll chaotic attractors, and its unique feature lies in the ability to control the number of scrolls. The investigation of the planar multi-scroll chaotic attractors and its dynamical behaviors is conducted through the analysis of phase plane portraits, bifurcation diagrams, and spectral entropies. The numerical simulations unveil a compelling relationship between the number of chaotic scrolls and specific control parameters governing the model. To further validate the findings, a 3D autonomous tabu learning single neuron model is implemented on a digital hardware platform. In an effort to extend the practical significance of this research, the multi-scroll chaotic phenomenon generated by the proposed model is deployed for image encryption. The fusion of mathematical modeling, digital hardware implementation, and practical application underscores the universality and significance of the proposed single neuron model in the fields of chaotic systems and engineering applications.

Natural Organic Honey-CNT Memristor Based Artificial Synaptic Devices for Sustainable Neuromorphic System

Neuromorphic computing is considered to have the potential to overcome the limitations of traditional von Neumann architecture due to its high efficiency, low energy consumption, and fault-tolerance. Hardware components that can emulate the synaptic plasticity of neurons, i.e. artificial synaptic devices, are required by neuromorphic systems. New devices have been examined for such components, such as phase-change artificial synapse, ferroelectric artificial synapse, and memristor synapses. Among them, memristor, a two-terminal metal-insulator-metal structure that are analogous to a biological synapse with presynaptic neuron (top electrode), postsynapticneuron (bottom electrode), and synaptic cleft (memristive film), is a promising device technology because of its tunable resistance, scalability, 3D integration compatibility, low power consumption, and relatively high speed. In contrary to inorganic materials such as metal oxides, natural organic materials have attracted interest to form the memristive layer because they are renewable, biodegradable, sustainable, biocompatible, and environmentally friendly. In this paper, honey solution embedded with carbon nanotubes (CNTs) was processed into the memristive layer by a low cost solution-based process, with synaptic plasticity of the final honey-CNT memristors characterized, including forget and relearn, spike-rate-dependent plasticity, spike-voltage-dependent plasticity, short-term to long-term memory transition, paired pulse facilitation, and spatial supra-linear summation behaviors. The successful emulation of these essential biological synaptic behaviors demonstrates the potential of honey-CNT memristors as a viable hardware component in neuromorphic computing systems.

Harnessing a WOx-based flexible transparent memristor synapse with a hafnium oxide layer for neuromorphic computing.

Transparent memristor-based neuromorphic synapses are expected to be specialised devices for high-speed information transmission and processing. The synaptic linearity and potentiation/depression cycles are imperative issues for the application of memristors. This work explores a memristor for improving switching uniformity by introducing a thin HfOx interfacial layer as a diffusion-limiting layer sandwiched between WOx and ITO bottom electrodes. An optimized HfOx thickness not only provides the best switching properties but also shows superior synaptic properties. The optimized 15 nm thin WOx layer can retain the memristor's excellence in P/D linearity, a cycling stability of 494 epochs and image recognition up to 3 mm bending, making it suitable for flexible devices. The artificial synapse is capable of reversible short-term and long-term learning behaviors confirmed by spike-timing-dependent-plasticity (STDP) results. X-ray photoelectron spectroscopy confirms the device composition and provides the oxygen vacancy concentration at the WOx/HfOx interface to realize the switching mechanism. The thicknesses of the different layers are estimated from the high-resolution transmission electron microscopy observations. The fabricated device exhibits 92.2% transparency, as confirmed by the UV-Vis spectrum.

Implementation of Multiple-Step Quantized STDP Based on Novel Memristive Synapses

Implementation of Multiple-Step Quantized STDP Based on Novel Memristive Synapses

Constructing Memristive Hindmarsh–Rose Neuron with Countless Coexisting Firings

Memristor synapses have been widely introduced into neuronal models to investigate the effects of external magnetic fields. However, there is a relative lack of research on the external-induced electric fields in neurons. In this paper, a 4D-memristive Hindmarsh–Rose neuron model is constructed by introducing a memristor and an electric field variable, which can generate complex neural firing. Notably, numerical simulations reveal that the initial conditions of the memristor can induce different firing patterns, exhibiting a unique fractal structure in the basin of attractions. Remarkably, the offset parameters of the internal variables of the neuron can be canceled out so that the offset boosting of the variables can be achieved according to the initial values, giving rise to an uncountably many hidden attractors with homogeneous multistability. This model provides the first example of generating uncountably many attractors in a memristive neuron model without relying on trigonometric functions, significantly advancing our understanding of neuronal dynamics. Finally, a digital circuit is designed and implemented on the RISC-V platform to verify the numerical simulation and theoretical analysis. The findings of this study have a certain implication for the development of advanced neuromorphic computing systems and the understanding of complex neuronal behaviors in the presence of external electric fields.

Investigation of Synchronization Control of the Memristive/Resistive-Coupled Neural Network with Noise Effect

This study comprehensively investigates synchronization control of the memristive- and resistive-coupled neural network with noise effect. In this context, a sub-network is constructed with three chaotic memristive Hindmarsh–Rose neurons coupled by the memristive synapses. Then, two equivalent sub-networks are linked by the diffusive resistive synapse under a distinctive global network structure including two sub-networks. Thus, a double (memristive/resistive) control mechanism is used to achieve the synchronization state. The first mechanism is the memristive synaptic coupling in each sub-network, and the second one is the diffusive resistive coupling between two sub-networks constructing the global network. The synchronization between the neurons is proven theoretically through the Lyapunov stability and Lipschitz theorems and this proof process is verified with numerical experiments. According to the theoretical and numerical results, by adjusting memristive/resistive synaptic coupling weights, a synchrony behavior is generated successfully in these sub-networks and global networks. Additionally, this paper also handles the Gaussian white noise effect on the global network to examine continuity capability and stability of synchronization. Two different frameworks are built by applying the noise with multiple noise power levels and periods to (i) one of the resistive coupling neurons in each sub-network directly and (ii) one neuron in a sub-network without direct contact with the other sub-network. With these two characteristic designs, it is proven that the second design provides more robust characteristics since it utilizes the double control mechanism, which enhances the stability throughout external interferences enabling the network to sustain synchronous states.

Grid multi-double-scroll attractors in a magnetized Hopfield neural network with a memristive self-connection synapse

Grid multi-double-scroll attractors in a magnetized Hopfield neural network with a memristive self-connection synapse

Energy consumption in the synchronization of neurons coupled by electrical or memristive synapse

The energy consumption in synapses has been an issue of great concern, and we apply the resistor to function as the electrical synapse, and the memristor is utilized as the chemical synapse. We investigate the dynamics of neurons during the synchronization, as well as the energy consumption associated with coupled channels. The numerical results revealed that two neurons with different initial values can get into a complete synchronization, regardless of whether they are coupled via resistive or memristive synapse. Furthermore, the energy consumption of electrical synaptic coupling is companied by an energy phase transition burst that causes a sharp increase or decrease in the channel energy consumption. In contrast, memristive coupling promotes steady information and energy exchange between neurons, with relatively lower energy consumption compared to electrical synaptic coupling. Meanwhile, the memconductance of the memristor is lower than that of resistive coupling, leading to a lower coupling strength when the synchronization is achieved. Furthermore, the employment of the energy switch to trigger synaptic activation or silence can effectively control synaptic activity, and neurons can spontaneously synchronize and maintain energy balance. The energy difference between chaotic neurons can cause the transitions between synchronization, desynchronization, and resynchronization. The findings may provide insights for developing genuine neural circuits and leveraging them in artificial neuron design.

Hardware software co-design for leveraging STDP in a memristive neuroprocessor

In neuromorphic computing, different learning mechanisms are being widely adopted to improve the performance of a specific application. Among these techniques, spike-timing-dependent plasticity (STDP) stands out as one of the most favored. STDP is simply managed by the temporal information of an event, which is biologically inspired. However, most of the prior works on STDP are focused on circuit implementation or software simulation for performance evaluation. Previous works also lack a comparative analysis of the performances of different STDP implementations. This study aims to provide a comprehensive assessment of STDP, centering on the performance across various applications such as classification (static and temporal datasets), control, and reservoir computing. Different applications necessitate distinct STDP configurations to achieve optimal performance with the neuroprocessor. Additionally, this work introduces an application-specific integrated circuit design of STDP circuitry. The design is based on current-controlled memristive synapse principles and utilizes 65 nm CMOS technology from IBM. The detailed presentation includes circuitry specifics, layout, and performance parameters such as energy consumption and design area.

Dynamic analysis and hardware implementation of multi-scroll Hopfield neural networks with three different memristor synapses

Dynamic analysis and hardware implementation of multi-scroll Hopfield neural networks with three different memristor synapses