Memristive Neuron Research Articles

Coherence resonance in a memristive map neuron and adaptive energy regulation

Involvement of memristive term and additive physical variables including magnetic flux and charge can enhance the physical description of the biophysical neurons. Neural circuits coupled with memristors can be built and tamed to mimic the intrinsic biophysical characteristics and dynamical properties of biological neurons, and these memristive oscillator models are effective in predicting the mode transition in neural activities and self-organization in collective electric behaviors of neural networks. Any proposal of memristive map neurons requires reliable physical description. For example, the energy definition and self-adaptive working mechanism are crucial to verify the reliability of memristive maps. A capacitive variable is useful to describe the membrane potential, while the complexity of ion channels requires careful evaluation and description by using inductive variables relative to the electromagnetic field. In this work, a charge-controlled memristor is connected to an inductor in series for building a hybrid ion channel, and then a capacitor and a nonlinear resistor are combined to couple the ion channel. As a result, a simple memristive neural circuit is designed to discern the inner effect of electric field and magnetic field synchronously. The energy function is defined and verified with theoretical proof. Furthermore, a linear transformation is applied to convert this memristive neuron into a memristive map with an exact energy description, in which its dynamics and mode transition will be controlled by an adaptive law when its energy is beyond the threshold. Additive noise is imposed to induce coherence resonance, which can be detected by using the statistical analysis and average value for Hamilton energy function during changes in noise intensity. This scheme provides guidance for energy definition in memristive maps and the intrinsic energy regulation mechanism in neural activities is explained from physical and dynamical aspects.

An image encryption scheme based on an improved memristive neuron chaotic system

With fast-developing Internet and communication, the security transmission of image in network has become a research highlight. So far, researchers have designed a lot of image encryption methods based on chaotic models, some of them are not secure enough. To enhance security of image transmission on the network, in this paper, an image encryption method is developed from a 3D memeristive FitzHugh-Nagumo (FHN) neuron. First, a 3D memeristive FHN model is obtained by connecting a memeristor into a 2D FHN model, and dynamics for 3D FHN model are estimated by applying phase diagrams, bifurcation and Lyapunov exponent. Then, an image encryption algorithm is proposed by using this 3D model. Finally, security of encryption algorithm is estimated. Simulation results confirm the effectiveness of encryption scheme.

Design of the non‐associative learning memristive circuit and the application to brightness adaptation

AbstractNon‐associative learning plays an important role in the biological and cognitive systems and is one of the important pathways for unsupervised learning in biology. The proposed non‐associative learning memristive circuit is composed of a synapse module, a memristive neuron module, and a response boosting module. The circuit can simulate the characteristics of habituation and sensitization, which attenuates the output response when receiving repeated input stimuli and augments the output response when receiving strong input stimuli, and the circuit also has a certain anti‐interference ability. Additionally, the designed circuit is suitable for adaptive replication experiments in visual systems and automatic exposure compensation systems, demonstrating the practicality of the circuit.

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.

Enhance controllability of a memristive neuron under magnetic field and circuit approach

Enhance controllability of a memristive neuron under magnetic field and circuit approach

A tantalum oxide based memristive neuron device for anomaly detection application

Anomaly detection, a data intensive task, is very important in wide application scenarios. Memristor has shown excellent performance in data intensive tasks. However, memristor used for anomaly detection has rarely been reported. In this Letter, a tantalum oxide (TaOx) memristive neuron device has been developed for anomaly detection application. TaOx, a CMOS compatible material, based memristor shows reliable threshold switching characteristics, which is suitable for constructing memristive neuron. Furthermore, the output frequency of the memristive neuron is found to be proportionate to the applied stimulus intensity and at an inflection point starts to decrease, namely, thresholding effect. Based on the thresholding effect of the neuron output, the application of the memristive neuron for anomaly detection has been simulated. The results indicate that the TaOx memristive neuron with thresholding effect shows better performance (98.78%) than the neuron without threshoding effect (90.89%) for anomaly detection task. This work provided an effective idea for developing memristive anomaly detection system.

A multifunctional and reconfigurable memristive neuron circuit

AbstractA memristive neuron circuit which is capable of encoding input information via temporal and adaptive frequency methods within the same structure is proposed. Its multifunctionality allows switching between these modes by adjusting circuit capacitors. Furthermore, it is investigated how other parameters like input voltage and resistance influence output properties for different encoding methods. This research contributes to the development of more flexible neuromorphic computing systems.

Dynamics in a light-sensitive neuron with two capacitive variables

Involvement of two capacitive variables into neuron models provides better description of the cell membrane property and then the diversity effect of electromagnetic field inner and outer of the cell membrane can be estimated in clear way. Specific electric components can be combined to build equivalent neural circuits for reproducing similar neural activities under some self-adaptive control schemes. A phototube converts external light into electric stimuli and the injected energy is encoded to excite the cell membranes for presenting suitable firing patterns. Two capacitors are connected via a linear resistor for mimicking the energy exchange and changes of membrane potentials. Combination of memristor into an additive branch circuit of the neural circuit can estimate the effect of electromagnetic induction and energy absorption. The energy function H for this light-sensitive and memristive neuron is calculated in theoretical way, and the average energy function 〈H〉 can predict the occurrence of stochastic resonance, which can be confirmed by estimating the distribution of signal to noise ratios. The firing mode is relative to the energy value of the neuron, and a control law is suggested to control the mode transition in neural activities in an adaptive way.

A joint image encryption based on a memristive Rulkov neuron with controllable multistability and compressive sensing

Image encryption, as a critical branch, has attracted increasing attention to the demand for information security specifically in the era of artificial intelligence (AI). A chaotic sequence is regarded as an important encryption source, and compressive sensing provides an effective technology for obtaining and reconstructing sparse or compressible signals in applied electronic engineering. In this work, the detouring matching pursuit algorithm and DNA coding are utilized to increase the performance based on a newly developed chaotic firing neuron. A memristor as the electromagnetic component is proven to enhance synaptic plasticity and emulate the synaptic connections in the brain. A unique discrete memristive neuron is derived for exploring the dynamics of neuron firing. By modifying the feedback from the memristor various coexisting neuronal chaotic firing are possessed. Because of the periodic evolution of the resistor from the memristor, the derived memristive Rulkov neuron exhibits coexisting homogeneous and heterogeneous multistability, which enables amplitude controllability and different types of coexisting chaotic firings. Circuit implementation based on CH32 is built to verify the controllability of the coexisting dynamics.

Energy and self-adaption in a memristive map neuron

Both capacitor and charge-controlled memristor (CCM) can regulate the inner energy levels during the release and accumulation of charges in these electronic components. Combination of inductive and capacitive components is crucial to build an oscillatory circuit, and incorporation of a nonlinear resistor is helpful to support nonlinear oscillation under continuous energy exchange between magnetic field and electric field. Relation between physical variables in neural circuits can be described by some equivalent nonlinear oscillators by applying scale transformation on the variables and parameters for the neural circuit, and the energy function becomes available from theoretical approach. In this work, a simple neural circuit coupled with a CCM is designed to obtain a memristive neuron model, and then linear transformation is applied to the sampled variables involved with time scale for time step, a new memristive map is obtained to discuss its energy characteristic and self-controllability under energy flow. The result provides guidance to measure the reliability of maps, and the new derived maps with reliable physical evidences can be further used in digital signal processing.

A discrete memristive neuron and its adaptive dynamics

A discrete memristive neuron and its adaptive dynamics

Bifurcations to bursting oscillations in memristor-based FitzHugh-Nagumo circuit

To characterize neuronal firing activities and elucidate its bifurcation mechanisms, a memristor-based FitzHugh-Nagumo (FHN) circuit is designed based on the FHN circuit architecture combined with a first-order memristive simulator, and its normalized system with periodic and quasi-periodic bursting oscillations is established. With the change of externally applied excitation, the memristor-based FHN system has the time-varying equilibrium point where the number, position and stability evolve slowly over time. In an evolution period of the time-domain waveform, different fold and/or Hopf bifurcations are triggered, resulting in periodic or quasi-periodic bursting oscillations. To explain the intrinsic bifurcation mechanisms, the fold and Hopf bifurcation sets are built and the transitions between the resting and spiking states are demonstrated, thus identifying the Hopf/fold and Hopf/Hopf bursting oscillations. Finally, based on the circuit simulation model, analog circuit simulations and hardware circuit measurements are developed for the memristor-based FHN circuit to confirm MATLAB numerical simulations. In addition, it is worth noting that the proposed circuit is a simple non-autonomous memristive neuron circuit that is particularly easy to physically implement.

Generating multi-scroll chaotic attractor in a three-dimensional memristive neuron model

The tabu learning single neuron model with external stimulus is a significant model for describing the firing activities of neurons. The current generated by a magnetic-controlled memristor has the properties of electromagnetic induction current, which can be taken as special external interference in a complex physical environment. Can this neuron model still exhibit complex and diverse firing activities when memristive current is used instead of external stimulus? To solve this puzzle, this article proposes a novel three-dimensional (3-D) memristive tabu learning neuron (MTLN) model by substituting external stimulus with memristive current, which can generate a multi-scroll chaotic attractor (MSCA) whose scrolls grow with time. The MTLN model has a line equilibrium set with different stability types relied on the initial position of the memristor internal state. Using some numerical methods, complex dynamical behaviors and scroll-growing MSCAs are uncovered. Further, a simple scroll-control scheme is designed and its effectiveness is demonstrated numerically and theoretically. The results manifest that the scrolls of the generated MSCAs can grow with time and can be also controlled by bounding the memristor internal state. In addition, an FPGA hardware platform is exploited for the 3-D MTLN model and the scroll-growing and scroll-controlled MSCAs are yielded experimentally to confirm the numerical results.

A novel memristive neuron model and its energy characteristics.

The functional neurons are basic building blocks of the nervous system and are responsible for transmitting information between different parts of the body. However, it is less known about the interaction between the neuron and the field. In this work, we propose a novel functional neuron by introducing a flux-controlled memristor into the FitzHugh-Nagumo neuron model, and the field effect is estimated by the memristor. We investigate the dynamics and energy characteristics of the neuron, and the stochastic resonance is also considered by applying the additive Gaussian noise. The intrinsic energy of the neuron is enlarged after introducing the memristor. Moreover, the energy of the periodic oscillation is larger than that of the adjacent chaotic oscillation with the changing of memristor-related parameters, and same results is obtained by varying stimuli-related parameters. In addition, the energy is proved to be another effective method to estimate stochastic resonance and inverse stochastic resonance. Furthermore, the analog implementation is achieved for the physical realization of the neuron. These results shed lights on the understanding of the firing mechanism for neurons detecting electromagnetic field.

Light-controlled threshold switching memristive neuron devices for color photoreceptor

Light-controlled threshold switching memristive neuron devices for color photoreceptor

A stochastic photo-responsive memristive neuron for an in-sensor visual system based on a restricted Boltzmann machine.

In-sensor computing has gained attention as a solution to overcome the von Neumann computing bottlenecks inherent in conventional sensory systems. This attention is due to the ability of sensor elements to directly extract meaningful information from external signals, thereby simplifying complex data. The advantage of in-sensor computing can be maximized with the sampling principle of a restricted Boltzmann machine (RBM) to extract significant features. In this study, a stochastic photo-responsive neuron is developed using a TiN/In-Ga-Zn-O/TiN optoelectronic memristor and an Ag/HfO2/Pt threshold-switching memristor, which can be configured as an input neuron in an in-sensor RBM. It demonstrates a sigmoidal switching probability depending on light intensity. The stochastic properties allow for the simultaneous exploration of various neuron states within the network, making identifying optimal features in complex images easier. Based on semi-empirical simulations, high recognition accuracies of 90.9% and 95.5% are achieved using handwritten digit and face image datasets, respectively. In addition, the in-sensor RBM effectively reconstructs abnormal face images, indicating that integrating in-sensor computing with probabilistic neural networks can lead to reliable and efficient image recognition under unpredictable real-world conditions.

Energy controls wave propagation in a neural network with spatial stimuli

Nervous system has distinct anisotropy and some intrinsic biophysical properties enable neurons present various firing modes in neural activities. In presence of realistic electromagnetic fields, non-uniform radiation activates these neurons with energy diversity. By using a feasible model, energy function is obtained to predict the growth of synaptic connections of these neurons. Distribution of average value of the Hamilton energy function vs. intensity of noisy disturbance can predict the occurrence of coherence resonance, which the neural activities show high regularity by applying noisy disturbance with moderate intensity. From physical viewpoint, the average energy value has similar role average power for the neuron. Non-uniform spatial disturbance is applied and energy is injected into the neural network, statistical synchronization factor is calculated to predict the network synchronization stability and wave propagation. The intensity for field coupling is adaptively controlled by energy diversity between adjacent neurons. Local energy balance will terminate further growth of the coupling intensity; otherwise, heterogeneity is formed in the network due to energy diversity. Furthermore, memristive channel current is introduced into the neuron model for perceiving the effect of electromagnetic induction and radiation, and a memristive neuron is obtained. The circuit implement of memristive circuit depends on the connection to a magnetic flux-controlled memristor into the mentioned neural circuit in an additive branch circuit. The connection and activation of this memristive neural network are controlled under external spatial electromagnetic radiation by capturing enough field energy. Continuous energy collection and exchange generate energy diversity and synaptic connection is created to regulate the synchronous firing patterns and energy balance.

Dynamics in a memristive neuron under an electromagnetic field

Dynamics in a memristive neuron under an electromagnetic field

FPGA implementation of memristive Hindmarsh–Rose neuron model: Low cost and high-performing through hybrid approximation

Recent memristive neuron models have improved the ability of researchers to describe the operation of biological neurons. We employ FPGA implementation for a memristive neuron model, FPGA implementation provides an improved hardware system for mimicking brain behavior, and provides a better understanding of neurological phenomena. In this manuscript, a memristive Hindmarsh–Rose neuron model taking into account the influence of electromagnetic radiation and two multiplierless approximations of this model are presented. These approximations are piecewise linear (PWL) model and hybrid model, which mimic the behaviors of the original model. In the hybrid model, power 2-based approximation and a novel approximation are used simultaneously. The proposed models were subjected to error analysis, revealing that the PWL model exhibited mean NRMSEs of 3.56% and 2.3% for bursting and spiking modes, respectively. On the other hand, the hybrid model displayed mean NRMSEs of 1.92% and 1.34% for bursting and spiking modes. According to the FPGA implementation results, the PWL model yields the overall resources saving of 75.82% and 47.8% speed-up. But in the hybrid model, by optimizing the approximation functions, in addition to obtaining higher accuracy, better results are provided with a 67.3% improvement in the maximum frequency and the 78.64% saving in hardware resources.

A New Compact Model for Third-Order Memristive Neuron With Box-Shaped Hysteresis and Dynamics Analysis

This paper proposes a new compact model and presents circuit-theoretical analysis for a third-order memristive neuromorphic element fabricated by S. Kumar et al. The proposed model mainly consists of a box-shaped resistor model and a simplified piecewise-linear memristor model. Since the dynamic behavior of the box-shaped hysteresis in the quasi-static current-voltage curve is mostly unexplored, we firstly extract the box-shaped resistor model and construct its oscillators. The coordinate system shifting method and dynamic route analysis method are used to reveal the operating mechanism of box-shaped resistor-based oscillators. Both the theoretical analysis and simulation verification indicate that the box-shaped hysteresis characteristic facilitates the generation of neuromorphic action potentials. The proposed new compact model not only captures quasi-static characteristics, but also includes dynamic behaviors, such as action potential, periodic spiking, and periodic bursting. The influences of the model parameters are further investigated to reveal the mechanism of the neuromorphic behaviors in the box-shaped hysteresis and positive differential resistance regions. Simulation results manifest the feasibility of the proposed model and the correctness of the presented analysis methods, which pave the way to the optimized design of memristive devices and the research of neuromorphic dynamics.