Robust Spike-Based Continual Meta-Learning Improved by Restricted Minimum Error Entropy Criterion

The spiking neural network (SNN) is regarded as a promising candidate to deal with the great challenges presented by current machine learning techniques, including the high energy consumption induced by deep neural networks. However, there is still a great gap between SNNs and the online meta-learning performance of artificial neural networks. Importantly, existing spike-based online meta-learning models do not target the robust learning based on spatio-temporal dynamics and superior machine learning theory. In this invited article, we propose a novel spike-based framework with minimum error entropy, called MeMEE, using the entropy theory to establish the gradient-based online meta-learning scheme in a recurrent SNN architecture. We examine the performance based on various types of tasks, including autonomous navigation and the working memory test. The experimental results show that the proposed MeMEE model can effectively improve the accuracy and the robustness of the spike-based meta-learning performance. More importantly, the proposed MeMEE model emphasizes the application of the modern information theoretic learning approach on the state-of-the-art spike-based learning algorithms. Therefore, in this invited paper, we provide new perspectives for further integration of advanced information theory in machine learning to improve the learning performance of SNNs, which could be of great merit to applied developments with spike-based neuromorphic systems.     

Dynamical origin of independent spiking and bursting activity in neural microcircuits

The relationship between spiking and bursting dynamics is a key question in neuroscience, particularly in understanding the origins of different neural coding strategies and the mechanisms of motor command generation and neural circuit coordination. Experiments indicate that spiking and bursting dynamics can be independent. We hypothesize that different mechanisms for spike and burst generation, intrinsic neuron dynamics for spiking and a modulational network instability for bursting, are the origin of this independence. We tested the hypothesis in a detailed dynamical analysis of a minimal inhibitory neural microcircuit (motif) of three reciprocally connected Hodgkin-Huxley neurons. We reduced this high-dimensional dynamical system to a rate model and showed that both systems have identical bifurcations from tonic spiking to burst generation, which, therefore, does not depend on the details of spiking activity.     

一类自突触作用下神经元电路的设计和模拟

神经元在自突触作用下可以诱发各类放电活动的迁移, 神经元动作电位对电自突触的响应比较敏感. 通常用包含延迟因子和增益的反馈回路电流来刻画自突触对神经元动作电位的影响. 基于Pspice软件, 设计了包含自突触效应的神经元电路, 用以延迟反馈电路来模拟电自突触对电位的调制作用. 研究结果发现: 1)在外界刺激和电自突触回路协同作用下, 神经元电路输出信号可以呈现静息态, 尖峰放电, 簇放电状态. 2)在时变增大的外界刺激下和自突触回路驱动下, 神经元电路的输出电位序列在多种电活动模式之间(静息, 尖峰放电, 簇放电)交替出现, 其机理在于自突触回路具有记忆特性, 神经元对于不同的外界刺激可以做出不同模式的响应. 3)在给定比较大外界刺激下, 改变反馈回路的增益, 发现电路输出的序列也可以呈现不同模式交替, 即神经元对于相同的刺激可以通过自我调节自突触增益来产生不同模式的响应, 其机理可能在于回路的有效反馈, 这有助于理解突触的可塑性.     

Thermal impact on spiking properties in Hodgkin-Huxley neuron with synaptic stimulus

The effect of environmental temperature on neuronal spiking behaviors is investigated by numerically simulating the temperature dependence of spiking threshold of the Hodgkin-Huxley neuron subject to synaptic stimulus. We find that the spiking threshold exhibits a global minimum in a specific temperature range where spike initiation needs weakest synaptic strength, which form the engineering perspective indicates the occurrence of optimal use of synaptic transmission in the nervous system. We further explore the biophysical origin of this phenomenon associated with ion channel gating kinetics and also discuss its possible biological relevance in information processing in neuronal systems.      

基于忆阻器的spiking神经网络在图像边缘提取中的应用

根据生物视觉系统的功能原理,用忆阻器模拟生物突触,结合忆阻器的记忆特性和spiking神经网络的高效处理能力,构造了一种可用于图像边缘提取的三层spiking神经网络模型,该网络用忆阻器电导的变化量来表征图像边缘信息,仿真结果表明,该方法的边缘提取结果具有连续性、光滑性、低误检漏检性和边缘定位准确性,该神经网络的处理过程符合生物信息处理机制,为视觉系统的仿生实现提供了新的思路。     

A Fault Diagnosis Method Considering Meteorological Factors for Transmission Networks Based on P Systems

Bad meteorological conditions may reduce the reliability of power communication equipment, which can increase the distortion possibility of fault information in the communication process, hence raising its uncertainty and incompleteness. To address the issue, this paper proposes a fault diagnosis method for transmission networks considering meteorological factors. Firstly, a spiking neural P system considering a meteorological living environment and its matrix reasoning algorithm are designed. Secondly, based on the topology structure of the target power transmission network and the action logic of its protection devices, a diagnosis model based on the spiking neural P system considering the meteorological living environment is built for each suspicious fault transmission line. Following this, the action messages of protection devices and corresponding temporal order information are used to obtain initial pulse values of input neurons of the diagnosis model, which are then modified with the gray fuzzy theory. Finally, the matrix reasoning algorithm of each model is executed in a parallel manner to obtain diagnosis results. Experiment results achieved out on IEEE 39-bus system show the feasibility and effectiveness of the proposed method.     

Characterizing neural activities evoked by manual acupuncture through spiking irregularity measures

The neural system characterizes information in external stimulations by different spiking patterns. In order to examine how neural spiking patterns are related to acupuncture manipulations, experiments are designed in such a way that different types of manual acupuncture （MA） manipulations are taken at the ＇Zusanli＇ point of experimental rats, and the induced electrical signals in the spinal dorsal root ganglion are detected and recorded. The interspike interval （ISI） statistical histogram is fitted by the gamma distribution, which has two parameters： one is the time-dependent firing rate and the other is a shape parameter characterizing the spiking irregularities. The shape parameter is the measure of spiking irregularities and can be used to identify the type of MA manipulations. The coefficient of variation is mostly used to measure the spike time irregularity, but it overestimates the irregularity in the case of pronounced firing rate changes. However, experiments show that each acupuncture manipulation will lead to changes in the firing rate. So we combine four relatively rate- independent measures to study the irregularity of spike trains evoked by different types of MA manipulations. Results suggest that the MA manipulations possess unique spiking statistics and characteristics and can be distinguished according to the spiking irregularity measures. These studies have offered new insights into the coding processes and information transfer of acupuncture.     

Pattern formation and encoding rhythms analysis on a spiking/bursting neuronal network

In this work we study the formation of patterns of neuronal activity when some input are presented to the network. For this task a recently developed model of neuron is utilized. This model requires a very low computational effort but presents many characteristics of more complex models such as, spiking, bursting and sub-threshold oscillations, and therefore the realistic study of the behavior of big ensembles of neurons can be aborded, even under real time conditions. New results of the application of the wavelet transform technique to the analysis of pattern formation and the possible encoding of rhythms are presented; they show that this simple, low-computational, neuron model behaves much like more complex ones.     

Channel block-optimized spiking activity of Hodgkin-Huxley neurons on random networks

In this paper, we numerically study the effect of channel block on the spiking temporal coherence and spatial synchronization on Hodgkin-Huxley (HH) neuron networks. It is found that under sodium CB the spike coherence is badly reduced, and the synchronization can, depending on the network randomness (the fraction of random shortcuts), be either enhanced or reduced, while, under potassium CB, the spike coherence can be enhanced but the synchronization is reduced. Interestingly, for certain networks of relatively large randomness, the neuron firings can achieve the best temporal coherence at an optimal potassium CB. These results show that under certain conditions channel blocking can increase and optimize the spike coherence and the synchronization on the complex HH neuron networks, whereby the neurons would exhibit a better and the best sub-threshold signal encoding.     

Variability of spatio-temporal patterns in non-homogeneous rings of spiking neurons

We show that a ring of unidirectionally delay-coupled spiking neurons may possess a multitude of stable spiking patterns and provide a constructive algorithm for generating a desired spiking pattern. More specifically, for a given time-periodic pattern, in which each neuron fires once within the pattern period at a predefined time moment, we provide the coupling delays and/or coupling strengths leading to this particular pattern. The considered homogeneous networks demonstrate a great multistability of various travelling time- and space-periodic waves which can propagate either along the direction of coupling or in opposite direction. Such a multistability significantly enhances the variability of possible spatio-temporal patterns and potentially increases the coding capability of oscillatory neuronal loops. We illustrate our results using FitzHugh-Nagumo neurons interacting via excitatory chemical synapses as well as limit-cycle oscillators.     

Modulating 3D memristor synapse by analog spiking pulses for bioinspired neuromorphic computing

Memristor based artificial synapses have demonstrated great potential for bioinspired neuromorphic computing in recent years. To emulate synaptic functions, such as short-term plasticity and long-term potentiation/depression, square pulses or combined complex pulse groups are applied on the device. However, in biological neuron systems, the action potentials are analog pulses with similar amplitudes. Furthermore, in biological systems, the intensity of the stimulus is coded into the frequency of action potentials to modulate the weight of synapses. Toward this programming method, we applied a series of analog spiking pulses with same peaks on Ru/TiO _x /TiN 3D memristor to emulate synaptic functions, such as long-term potentiation/depression and synaptic saturation. Moreover, we demonstrated the conductance change of the device under different stimulus frequencies of analog spiking pulses and described the statistical results of conductance change value, which shows that the device conductance has a larger change value under a higher spiking frequency with identical pulse number. These results show that the analog spiking pulses can well modulate the memristor-based synaptic weight and have a great potential for bioinspired computing in the future.     

兴奋性和抑制性自反馈压制靠近Hopf分岔的神经电活动比较

突触输入刺激神经元产生的电活动,在神经编码中发挥着重要作用.通常认为,兴奋性输入增强电活动,抑制性输入压制电活动.本文选取可调节电流衰减速度的突触模型,研究了兴奋性自突触在亚临界Hopf分岔附近压制神经元电活动的反常作用,与抑制性自突触的压制作用进行了比较,并采用相位响应曲线和相平面分析解释了压制作用的机制.对于单稳的峰放电,快速和中速衰减的兴奋性自突触分别可以诱发频率降低的峰放电和混合振荡(峰放电与阈下振荡的交替),而中速和慢速衰减的抑制性自突触也可以分别诱发频率降低的峰放电和混合振荡.对于与静息共存的峰放电,除上述两种行为外,中速衰减的兴奋性和慢速衰减的抑制性自突触还可以诱发静息.兴奋性和抑制性自突触电流在不同的衰减速度下,分别作用在峰放电的不同相位,才能诱发同类压制行为.结果丰富了兴奋性突触压制电活动反常作用的实例,获得了兴奋性和抑制性自突触压制作用机制的不同,给出了调控神经放电的新手段.     

Relation between single neuron and population spiking statistics and effects on network activity

To simplify theoretical analyses of neural networks, individual neurons are often modeled as Poisson processes. An implicit assumption is that even if the spiking activity of each neuron is non-Poissonian, the composite activity obtained by summing many spike trains limits to a Poisson process. Here, we show analytically and through simulations that this assumption is invalid. Moreover, we show with Fokker-Planck equations that the behavior of feedforward networks is reproduced accurately only if the tendency of neurons to fire periodically is incorporated by using colored noise whose autocorrelation has a negative component.     

Spiking sychronization regulated by noise in three types of Hodgkin—Huxley neuronal networks

In this paper,we study spiking synchronization in three different types of Hodgkin-Huxley neuronal networks,which are the small-world,regular,and random neuronal networks.All the neurons are subjected to subthreshold stimulus and external noise.It is found that in each of all the neuronal networks there is an optimal strength of noise to induce the maximal spiking synchronization.We further demonstrate that in each of the neuronal networks there is a range of synaptic conductance to induce the effect that an optimal strength of noise maximizes the spiking synchronization.Only when the magnitude of the synaptic conductance is moderate,will the effect be considerable.However,if the synaptic conductance is small or large,the effect vanishes.As the connections between neurons increase,the synaptic conductance to maximize the effect decreases.Therefore,we show quantitatively that the noise-induced maximal synchronization in the Hodgkin-Huxley neuronal network is a general effect,regardless of the specific type of neuronal network.     

Model architecture for associative memory in a neural network of spiking neurons

A synaptic connectivity model is assembled on a spiking neuron network aiming to build up a dynamic pattern recognition system. The connection architecture includes gap junctions and both inhibitory and excitatory chemical synapses based on Hebb’s hypothesis. The network evolution resulting from external stimulus is sampled in a properly defined frequency space. Neurons’ responses to different current injections are mapped onto a subspace using Principal Component Analysis. Departing from the base attractor, related to a quiescent state, different external stimuli drive the network to different fixed points through specific trajectories in this subspace.     

计算相位响应曲线的方波扰动直接算法

通过相位响应曲线可对具有极限环周期运动的动力系统的性质有更为深入的理解.神经元是一个典型的动力系统,因此相位响应曲线提供了一种研究神经元重复周期放电行为的新思路.本文提出一种求解相位响应曲线的方法,即方波扰动的直接算法,通过Hodgkin-Huxley,Fitz Hugh-Nagumo,Morris-Lecar和Hindmarsh-Rose神经元模型验证该算法可计算周期峰放电、周期簇放电的相位响应曲线.该算法克服了其他算法在运用过程中的局限性.利用该算法计算结果表明:周期峰放电的相位响应曲线类型是由其分岔类型所决定;在Morris-Lecar模型中发现一种开始于Hopf分岔终止于鞍点同宿轨道分岔的阈上周期振荡,其相位响应曲线属于第二类型.通过大量的相位响应曲线的计算发现相位响应的相对大小及正负性仅取决于扰动所施加的时间,而且周期簇放电的相位响应曲线比周期峰放电的相位响应曲线更为复杂.     

Diverse population-bursting modes of adapting spiking neurons

We study the dynamics of a noisy network of spiking neurons with spike-frequency adaptation (SFA), using a mean-field approach, in terms of a two-dimensional Fokker-Planck equation for the membrane potential of the neurons and the calcium concentration gating SFA. The long time scales of SFA allow us to use an adiabatic approximation and to describe the network as an effective nonlinear two-dimensional system. The phase diagram is computed for varying levels of SFA and synaptic coupling. Two different population-bursting regimes emerge, depending on the level of SFA in networks with noisy emission rate, due to the finite number of neurons.     

Parameter-dependent synchronization transition of coupled neurons with co-existing spiking and bursting

A firing pattern transition is simulated in the Leech neuron model, firstly from bursting to co-existence of spiking and bursting and then to spiking. The attraction domain of spiking and bursting for three different parameter values are calculated. Synchronization transition processes of two coupled Leech neurons, one is bursting and the other the co-existing spiking, are simulated for the three parameters. The three synchronization processes appear similar as the coupling strength increases, beginning from non-synchronization to complete synchronization through a complex dynamical procedure, but their detailed processes are different depending on the parameter values. The transition procedure is complex and the complete synchronization is in bursting for larger parameter values, while the process is simple with complete synchronization of spiking for smaller values. The potential relationship between complete synchronization and the attraction domain is also discussed. The results are instructive to understanding the synchronization behaviors of the coupled neuronal system with co-existing attractors.     

Negative self-feedback induced enhancement and transition of spiking activity for class-3 excitability

Post-inhibitory rebound (PIR) spike, which has been widely observed in diverse nervous systems with different physiological functions and simulated in theoretical models with class-2 excitability, presents a counterintuitive nonlinear phenomenon in that the inhibitory effect can facilitate neural firing behavior. In this study, a PIR spike induced by inhibitory stimulation from the resting state corresponding to class-3 excitability that is not related to bifurcation is simulated in the Morris-Lecar neuron. Additionally, the inhibitory self-feedback mediated by an autapse with time delay can evoke tonic/repetitive spiking from phasic/transient spiking. The dynamical mechanism for the PIR spike and the tonic/repetitive spiking is acquired with the phase plane analysis and the shape of the quasi-separatrix curve. The result extends the counterintuitive phenomenon induced by inhibition to class-3 excitability, which presents a potential function of inhibitory autapse and class-3 neuron in many neuronal systems such as the auditory system.     

Propagation of spiking regularity and double coherence resonance in feedforward networks

We investigate the propagation of spiking regularity in noisy feedforward networks (FFNs) based on FitzHugh-Nagumo neuron model systematically. It is found that noise could modulate the transmission of firing rate and spiking regularity. Noise-induced synchronization and synfire-enhanced coherence resonance are also observed when signals propagate in noisy multilayer networks. It is interesting that double coherence resonance (DCR) with the combination of synaptic input correlation and noise intensity is finally attained after the processing layer by layer in FFNs. Furthermore, inhibitory connections also play essential roles in shaping DCR phenomena. Several properties of the neuronal network such as noise intensity, correlation of synaptic inputs, and inhibitory connections can serve as control parameters in modulating both rate coding and the order of temporal coding.