Dynamical behaviors in discrete memristor-coupled small-world neuronal networks

The brain is a complex network system in which a large number of neurons are widely connected to each other and transmit signals to each other.The memory characteristic of memristors makes them suitable for simulating neuronal synapses with plasticity.In this paper,a memristor is used to simulate a synapse,a discrete small-world neuronal network is constructed based on Rulkov neurons and its dynamical behavior is explored.We explore the influence of system parameters on the dynamical behaviors of the discrete small-world network,and the system shows a variety of firing patterns such as spiking firing and triangular burst firing when the neuronal parameterαis changed.The results of a numerical simulation based on Matlab show that the network topology can affect the synchronous firing behavior of the neuronal network,and the higher the reconnection probability and number of the nearest neurons,the more significant the synchronization state of the neurons.In addition,by increasing the coupling strength of memristor synapses,synchronization performance is promoted.The results of this paper can boost research into complex neuronal networks coupled with memristor synapses and further promote the development of neuroscience.

Dynamics and synchronization in a memristor-coupled discrete heterogeneous neuron network considering noise

Research on discrete memristor-based neural networks has received much attention.However,current research mainly focuses on memristor–based discrete homogeneous neuron networks,while memristor-coupled discrete heterogeneous neuron networks are rarely reported.In this study,a new four-stable discrete locally active memristor is proposed and its nonvolatile and locally active properties are verified by its power-off plot and DC V–I diagram.Based on two-dimensional(2D)discrete Izhikevich neuron and 2D discrete Chialvo neuron,a heterogeneous discrete neuron network is constructed by using the proposed discrete memristor as a coupling synapse connecting the two heterogeneous neurons.Considering the coupling strength as the control parameter,chaotic firing,periodic firing,and hyperchaotic firing patterns are revealed.In particular,multiple coexisting firing patterns are observed,which are induced by different initial values of the memristor.Phase synchronization between the two heterogeneous neurons is discussed and it is found that they can achieve perfect synchronous at large coupling strength.Furthermore,the effect of Gaussian white noise on synchronization behaviors is also explored.We demonstrate that the presence of noise not only leads to the transition of firing patterns,but also achieves the phase synchronization between two heterogeneous neurons under low coupling strength.

基于迁移QCNN的孪生网络轴承故障诊断方法

轴承故障诊断对于降低旋转机械的损坏风险,进一步提高经济效益具有重要意义;深度学习在轴承故障诊断中应用广泛,但是深度学习模型在训练与测试时容易受到噪声的干扰导致性能下降;并且轴承的工况变化频繁,不同工况下的数据采集困难;对此,提出了一种基于迁移QCNN的孪生网络轴承故障诊断方法,先预训练QCNN获取具有较强判别性的模型参数,将预训练的参数迁移到QCNN作为子网络的孪生网络中,然后正常训练孪生网络获取模型,最后将测试数据与故障数据组成数据对输入模型,即可得到测试数据的故障类型;该方法将QCNN与孪生网络相结合,QCNN中的Quadratic神经元具有强大的特征提取能力,孪生网络共享权重和相对关系的训练方式,使得模型可以缓解噪声和工况数据不平衡问题的影响;实验结果显示,相较与传统机器学习模型和QCNN等模型,所提出方法在面对噪声和工况数据不平衡问题表现更好。

Scopolamine causes delirium-like brain network dysfunction and reversible cognitive impairment without neuronal loss

Delirium is a severe acute neuropsychiatric syndrome that commonly occurs in the elderly and is considered an independent risk factor for later dementia.However,given its inherent complexity,few animal models of delirium have been established and the mechanism underlying the onset of delirium remains elusive.Here,we conducted a comparison of three mouse models of delirium induced by clinically relevant risk factors,including anesthesia with surgery(AS),systemic inflammation,and neurotransmission modulation.We found that both bacterial lipopolysaccharide(LPS)and cholinergic receptor antagonist scopolamine(Scop)induction reduced neuronal activities in the delirium-related brain network,with the latter presenting a similar pattern of reduction as found in delirium patients.Consistently,Scop injection resulted in reversible cognitive impairment with hyperactive behavior.No loss of cholinergic neurons was found with treatment,but hippocampal synaptic functions were affected.These findings provide further clues regarding the mechanism underlying delirium onset and demonstrate the successful application of the Scop injection model in mimicking delirium-like phenotypes in mice.

面向SAR图像目标分类的CNN模型可视化方法

卷积神经网络(CNN)在合成孔径雷达(SAR)图像目标分类任务中应用广泛。由于网络工作机理不透明,CNN模型难以满足高可靠性实际应用的要求。类激活映射方法常用于可视化CNN模型的决策区域,但现有方法主要基于通道级或空间级类激活权重,且在SAR图像数据集上的应用仍处于起步阶段。基于此,该文从神经元特征提取能力和网络决策依据两个层面出发,提出了一种面向SAR图像的CNN模型可视化方法。首先,基于神经元的激活值,对神经元在其感受野范围内的目标结构学习能力进行可视化,然后提出一种通道-空间混合的类激活映射方法,通过对SAR图像中的重要区域进行定位,为模型的决策过程提供依据。实验结果表明,该方法给出了模型在不同设置下的可解释性分析,有效拓展了卷积神经网络在SAR图像上的可视化应用。

基于ANN模型的内冷型溶液除湿器性能研究

溶液除湿器因可被低品位热能驱动,且具有除湿效率高等优点而受到广泛关注,但其传质性能的预测准确度还有待提高.本文搭建了单通道内冷型溶液除湿实验平台,研究了不同参数对于除湿过程中传质性能的影响,同时,建立了基于MATLAB平台的人工神经网络(ANN)模型用于预测传质性能,并用上述实验数据对该ANN模型进行了验证.结果表明,ANN模型预测得出的Sh与实验Sh平均绝对相对偏差(MARD)为4.07%.与现有经验公式相比,建立的ANN模型预测精度更高.此外,还利用ANN模型研究了不同参数变化下的Sh的变化趋势,从而分析不同参数对除湿性能的影响.

一种Sigmoid为激活函数的高精度ANN电路的设计

提出了一款以Sigmoid为激活函数的人工神经网络(ANN)电路,实现对目标数据的分类,基于TSMC 180 nm CMOS工艺完成了电路设计、版图设计、仿真与流片测试,芯片面积为0.3375 mm^(2),1.8 V电源电压下电路功耗为216μW。3块芯片的测试结果显示,电路对目标数据的分类准确率为97.5%,误差在(-3.2%,3.27%)之间,平均误差为-1.51%。人工神经网络电路含有3个网络层,由15个Sigmoid激活函数神经元电路组成,两输入Sigmoid激活函数神经元电路由两个加权电路与一个Sigmoid激活函数电路组成。后仿真结果显示,在1.8 V电源电压下,神经元电路的功耗为8.02μW,最大噪声为√Hz,版图面积为0.001 mm^(2),最大相对误差为1.76%。

Advances in memristor based artificial neuron fabrication-materials,models,and applications

Spiking neural network(SNN),widely known as the third-generation neural network,has been frequently investigated due to its excellent spatiotemporal information processing capability,high biological plausibility,and low energy consumption characteristics.Analogous to the working mechanism of human brain,the SNN system transmits information through the spiking action of neurons.Therefore,artificial neurons are critical building blocks for constructing SNN in hardware.Memristors are drawing growing attention due to low consumption,high speed,and nonlinearity characteristics,which are recently introduced to mimic the functions of biological neurons.Researchers have proposed multifarious memristive materials including organic materials,inorganic materials,or even two-dimensional materials.Taking advantage of the unique electrical behavior of these materials,several neuron models are successfully implemented,such as Hodgkin–Huxley model,leaky integrate-and-fire model and integrate-and-fire model.In this review,the recent reports of artificial neurons based on memristive devices are discussed.In addition,we highlight the models and applications through combining artificial neuronal devices with sensors or other electronic devices.Finally,the future challenges and outlooks of memristor-based artificial neurons are discussed,and the development of hardware implementation of brain-like intelligence system based on SNN is also prospected.

One memristor–one electrolyte-gated transistor-based high energy-efficient dropout neuronal units

Artificial neural networks(ANN) have been extensively researched due to their significant energy-saving benefits.Hardware implementations of ANN with dropout function would be able to avoid the overfitting problem. This letter reports a dropout neuronal unit(1R1T-DNU) based on one memristor–one electrolyte-gated transistor with an ultralow energy consumption of 25 p J/spike. A dropout neural network is constructed based on such a device and has been verified by MNIST dataset, demonstrating high recognition accuracies(> 90%) within a large range of dropout probabilities up to40%. The running time can be reduced by increasing dropout probability without a significant loss in accuracy. Our results indicate the great potential of introducing such 1R1T-DNUs in full-hardware neural networks to enhance energy efficiency and to solve the overfitting problem.

The Lightweight Edge-Side Fault Diagnosis Approach Based on Spiking Neural Network

Network fault diagnosis methods play a vital role in maintaining network service quality and enhancing user experience as an integral component of intelligent network management.Considering the unique characteristics of edge networks,such as limited resources,complex network faults,and the need for high real-time performance,enhancing and optimizing existing network fault diagnosis methods is necessary.Therefore,this paper proposes the lightweight edge-side fault diagnosis approach based on a spiking neural network(LSNN).Firstly,we use the Izhikevich neurons model to replace the Leaky Integrate and Fire(LIF)neurons model in the LSNN model.Izhikevich neurons inherit the simplicity of LIF neurons but also possess richer behavioral characteristics and flexibility to handle diverse data inputs.Inspired by Fast Spiking Interneurons(FSIs)with a high-frequency firing pattern,we use the parameters of FSIs.Secondly,inspired by the connection mode based on spiking dynamics in the basal ganglia(BG)area of the brain,we propose the pruning approach based on the FSIs of the BG in LSNN to improve computational efficiency and reduce the demand for computing resources and energy consumption.Furthermore,we propose a multiple iterative Dynamic Spike Timing Dependent Plasticity(DSTDP)algorithm to enhance the accuracy of the LSNN model.Experiments on two server fault datasets demonstrate significant precision,recall,and F1 improvements across three diagnosis dimensions.Simultaneously,lightweight indicators such as Params and FLOPs significantly reduced,showcasing the LSNN’s advanced performance and model efficiency.To conclude,experiment results on a pair of datasets indicate that the LSNN model surpasses traditional models and achieves cutting-edge outcomes in network fault diagnosis tasks.

Effects of information transmission delay and channel blocking on synchronization in scale-free Hodgkin-Huxley neuronal networks

In this paper,we investigate the evolution of spatiotemporal patterns and synchronization transitions in dependence on the information transmission delay and ion channel blocking in scale-free neuronal networks.As the underlying model of neuronal dynamics,we use the Hodgkin-Huxley equations incorporating channel blocking and intrinsic noise.It is shown that delays play a significant yet subtle role in shaping the dynamics of neuronal networks.In particular,regions of irregular and regular propagating excitatory fronts related to the synchronization transitions appear intermittently as the delay increases.Moreover,the fraction of working sodium and potassium ion channels can also have a significant impact on the spatiotemporal dynamics of neuronal networks.As the fraction of blocked sodium channels increases,the frequency of excitatory events decreases,which in turn manifests as an increase in the neuronal synchrony that,however,is dysfunctional due to the virtual absence of large-amplitude excitations.Expectedly,we also show that larger coupling strengths improve synchronization irrespective of the information transmission delay and channel blocking.The presented results are also robust against the variation of the network size,thus providing insights that could facilitate understanding of the joint impact of ion channel blocking and information transmission delay on the spatiotemporal dynamics of neuronal networks.

Recognition system of leaf images based on neuronal network

In forest variety registration, visual traits of the plants appearance are widely used to discern different tree species. The new recognition system of leaf image strategy which based on neural network established to administrate a hierarchical list of leaf images, some sorts of edge detection can be performed to identify the individual tokens of every image and the frame of the leaf can be got to differentiate the tree species. An approach based on back-propagation neuronal network is proposed and the programming language for the implementation is also Riven by using Java. The numerical simulations results have shown that the proposed leaf strategt is effective and feasible.

Complete and phase synchronization in a heterogeneous small-world neuronal network

Synchronous firing of neurons is thought to be important for information communication in neuronal networks. This paper investigates the complete and phase synchronization in a heterogeneous small-world chaotic Hindmarsh Rose neuronal network. The effects of various network parameters on synchronization behaviour are discussed with some biological explanations. Complete synchronization of small-world neuronal networks is studied theoretically by the master stability function method. It is shown that the coupling strength necessary for complete or phase synchronization decreases with the neuron number, the node degree and the connection density are increased. The effect of heterogeneity of neuronal networks is also considered and it is found that the network heterogeneity has an adverse effect on synchrony.

Delay-aided stochastic multiresonances on scale-free FitzHugh-Nagumo neuronal networks

The stochastic resonance in paced time-delayed scale-free FitzHugh--Nagumo （FHN） neuronal networks is investigated. We show that an intermediate intensity of additive noise is able to optimally assist the pacemaker in imposing its rhythm on the whole ensemble. Furthermore, we reveal that appropriately tuned delays can induce stochastic multiresonances, appearing at every integer multiple of the pacemaker＇s oscillation period. We conclude that fine-tuned delay lengths and locally acting pacemakers are vital for ensuring optimal conditions for stochastic resonance on complex neuronal networks.

Improving Dendritic Neuron Model With Dynamic Scale-Free Network-Based Differential Evolution

Some recent research reports that a dendritic neuron model(DNM)can achieve better performance than traditional artificial neuron networks(ANNs)on classification,prediction,and other problems when its parameters are well-tuned by a learning algorithm.However,the back-propagation algorithm(BP),as a mostly used learning algorithm,intrinsically suffers from defects of slow convergence and easily dropping into local minima.Therefore,more and more research adopts non-BP learning algorithms to train ANNs.In this paper,a dynamic scale-free network-based differential evolution(DSNDE)is developed by considering the demands of convergent speed and the ability to jump out of local minima.The performance of a DSNDE trained DNM is tested on 14 benchmark datasets and a photovoltaic power forecasting problem.Nine meta-heuristic algorithms are applied into comparison,including the champion of the 2017 IEEE Congress on Evolutionary Computation(CEC2017)benchmark competition effective butterfly optimizer with covariance matrix adapted retreat phase(EBOwithCMAR).The experimental results reveal that DSNDE achieves better performance than its peers.

Spatial coherence resonance induced by coloured noise and parameter diversity in a neuronal network

Spatial coherence resonance in a two-dimensional neuronal network induced by additive Gaussian coloured noise and parameter diversity is studied. We focus on the ability of additive Gaussian coloured noise and parameter diversity to extract a particular spatial frequency （wave number） of excitatory waves in the excitable medium of this network. We show that there exists an intermediate noise level of the coloured noise and a particular value of diversity, where a characteristic spatial frequency of the system comes forth. Hereby, it is verified that spatial coherence resonance occurs in the studied model. Furthermore, we show that the optimal noise intensity for spatial coherence resonance decays exponentially with respect to the noise correlation time. Some explanations of the observed nonlinear phenomena are also presented.

Spiral Waves and Multiple Spatial Coherence Resonances Induced by Colored Noise in Neuronal Network

Gaussian colored noise induced spatial patterns and spatial coherence resonances in a square lattice neuronal network composed of Morris-Lecar neurons are studied.Each neuron is at resting state near a saddle-node bifurcation on invariant circle,coupled to its nearest neighbors by electronic coupling.Spiral waves with different structures and disordered spatial structures can be alternately induced within a large range of noise intensity.By calculating spatial structure function and signal-to-noise ratio(SNR),it is found that SNR values are higher when the spiral structures are simple and are lower when the spatial patterns are complex or disordered,respectively.SNR manifest multiple local maximal peaks,indicating that the colored noise can induce multiple spatial coherence resonances.The maximal SNR values decrease as the correlation time of the noise increases.These results not only provide an example of multiple resonances,but also show that Gaussian colored noise play constructive roles in neuronal network.

Plasticity-induced characteristic changes of pattern dynamics and the related phase transitions in small-world neuronal networks

Phase transitions widely exist in nature and occur when some control parameters are changed. In neural systems, their macroscopic states are represented by the activity states of neuron populations, and phase transitions between different activity states are closely related to corresponding functions in the brain. In particular, phase transitions to some rhythmic synchronous firing states play significant roles on diverse brain functions and disfunctions, such as encoding rhythmical external stimuli, epileptic seizure, etc. However, in previous studies, phase transitions in neuronal networks are almost driven by network parameters （e.g., external stimuli）, and there has been no investigation about the transitions between typical activity states of neuronal networks in a self-organized way by applying plastic connection weights. In this paper, we discuss phase transitions in electrically coupled and lattice-based small-world neuronal networks （LBSW networks） under spike-timing-dependent plasticity （STDP）. By applying STDP on all electrical synapses, various known and novel phase transitions could emerge in LBSW networks, particularly, the phenomenon of self-organized phase transitions （SOPTs）： repeated transitions between synchronous and asynchronous firing states. We further explore the mechanics generating SOPTs on the basis of synaptic weight dynamics.

Neuronal networks in mental diseases and neuropathic pain:Beyond brain derived neurotrophic factor and collapsin response mediator proteins

The brain is a complex network system that has the capacity to support emotion, thought, action, learning and memory, and is characterized by constant activity, constant structural remodeling, and constant attempt to compensate for this remodeling. The basic insight that emerges from complex network organization is that substantively different networks can share common key organizational principles. Moreover, the interdependence of network organization and behavior has been successfully demonstrated for several specific tasks. From this viewpoint, increasing experimental/clinical observations suggest that mental disorders are neural network disorders. On one hand, single psychiatric disorders arise from multiple, multifactorial molecular and cellular structural/functional alterations spreading throughout local/global circuits leading to multifaceted and heterogeneous clinical symptoms. On the other hand, various mental diseases may share functional deficits across the same neural circuit as reflected in the overlap of symptoms throughout clinical diagnoses. An integrated framework including experimental measures and clinical observations will be necessary to formulate a coherent and comprehensive understanding of how neural connectivity mediates and constraints the phenotypic expression of psychiatric disorders.

NNL:a domain-specific language for neural networks

Recent years,neural networks(NNs)have received increasing attention from both academia and industry.So far significant diversity among existing NNs as well as their hardware platforms makes NN programming a daunting task.In this paper,a domain-specific language(DSL)for NNs,neural network language(NNL)is proposed to deliver productivity of NN programming and portable performance of NN execution on different hardware platforms.The productivity and flexibility of NN programming are enabled by abstracting NNs as a directed graph of blocks.The language describes 4 representative and widely used NNs and runs them on 3 different hardware platforms(CPU,GPU and NN accelerator).Experimental results show that NNs written with the proposed language are,on average,14.5%better than the baseline implementations across these 3 platforms.Moreover,compared with the Caffe framework that specifically targets the GPU platform,the code can achieve similar performance.