Network Model Of Spiking Neurons Research Articles

Learning spiking neuronal networks with artificial neural networks: neural oscillations.

First-principles-based modelings have been extremely successful in providing crucial insights and predictions for complex biological functions and phenomena. However, they can be hard to build and expensive to simulate for complex living systems. On the other hand, modern data-driven methods thrive at modeling many types of high-dimensional and noisy data. Still, the training and interpretation of these data-driven models remain challenging. Here, we combine the two types of methods to model stochastic neuronal network oscillations. Specifically, we develop a class of artificial neural networks to provide faithful surrogates to the high-dimensional, nonlinear oscillatory dynamics produced by a spiking neuronal network model. Furthermore, when the training data set is enlarged within a range of parameter choices, the artificial neural networks become generalizable to these parameters, covering cases in distinctly different dynamical regimes. In all, our work opens a new avenue for modeling complex neuronal network dynamics with artificial neural networks.

Short-term neuronal and synaptic plasticity act in synergy for deviance detection in spiking networks.

Sensory areas of cortex respond more strongly to infrequent stimuli when these violate previously established regularities, a phenomenon known as deviance detection (DD). Previous modeling work has mainly attempted to explain DD on the basis of synaptic plasticity. However, a large fraction of cortical neurons also exhibit firing rate adaptation, an underexplored potential mechanism. Here, we investigate DD in a spiking neuronal network model with two types of short-term plasticity, fast synaptic short-term depression (STD) and slower threshold adaptation (TA). We probe the model with an oddball stimulation paradigm and assess DD by evaluating the network responses. We find that TA is sufficient to elicit DD. It achieves this by habituating neurons near the stimulation site that respond earliest to the frequently presented standard stimulus (local fatigue), which diminishes the response and promotes the recovery (global fatigue) of the wider network. Further, we find a synergy effect between STD and TA, where they interact with each other to achieve greater DD than the sum of their individual effects. We show that this synergy is caused by the local fatigue added by STD, which inhibits the global response to the frequently presented stimulus, allowing greater recovery of TA-mediated global fatigue and making the network more responsive to the deviant stimulus. Finally, we show that the magnitude of DD strongly depends on the timescale of stimulation. We conclude that highly predictable information can be encoded in strong local fatigue, which allows greater global recovery and subsequent heightened sensitivity for DD.

Next generation neural population models

Low-dimensional neural mass models are often invoked to model the coarse-grained activity of large populations of neurons and synapses and have been used to help understand the coordination of large scale brain rhythms. However, they are phenomenological in nature and, although motivated by neurobiological considerations, the absence of a direct link to an underlying biophysical reality is a weakness that means they may not be best suited to capturing some of the rich behaviors seen in real neuronal tissue. In this perspective article I discuss a simple spiking neuron network model that has recently been shown to admit to an exact mean-field description for synaptic interactions. This has many of the features of a neural mass model coupled to an additional dynamical equation that describes the evolution of population synchrony. Thisnext generationneural mass model is ideally suited to understanding the patterns of brain activity that are ubiquitously seen in neuroimaging recordings. Here I review the mean-field equations, the way in which population synchrony, firing rate, and average voltage are intertwined, together with their application in large scale brain modeling. As well as natural extensions of this new approach to modeling the dynamics of neuronal populations I discuss some of the open mathematical challenges in developing a statistical neurodynamics that can generalize the one discussed here.

Fault-Tolerant Spiking Neural Network Mapping Algorithm and Architecture to 3D-NoC-Based Neuromorphic Systems

Neuromorphic computing uses spiking neuron network models to solve machine learning problems in a more energy-efficient way when compared to conventional artificial neural networks. However, mapping the various network components to the neuromorphic hardware is not trivial to realize the desired model for an actual simulation. Moreover, neurons and synapses could be affected by noise due to external interference or random actions of other components (i.e., neurons), which eventually lead to unreliable results. This work proposes a fault-tolerant spiking neural network mapping algorithm and architecture to a 3D network-on-chip (NoC)-based neuromorphic system (R-NASH-II) based on a rank and selection mapping mechanism (RSM). The RSM allows the ranking and rapid selection of neurons for fault-tolerant mapping. Evaluation results show that with our proposed mechanism, we could maintain a mapping efficiency of 100% with 20% spare rate and a fault rate (40%) more than in the previous mapping framework. The Monte Carlo simulation evaluation of reliability shows that the RSM mechanism has increased the mean time to failure (MTTF) of the previous mapping technique by 43% on average. Furthermore, the operational availability of the RSM for mapping to a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$4\times 4\times 4$ </tex-math></inline-formula> (smallest) and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$6\times 6\times 6$ </tex-math></inline-formula> (largest) NoC is 88% and 67% respectively.

Training a spiking neuronal network model of visual-motor cortex to play a virtual racket-ball game using reinforcement learning.

Recent models of spiking neuronal networks have been trained to perform behaviors in static environments using a variety of learning rules, with varying degrees of biological realism. Most of these models have not been tested in dynamic visual environments where models must make predictions on future states and adjust their behavior accordingly. The models using these learning rules are often treated as black boxes, with little analysis on circuit architectures and learning mechanisms supporting optimal performance. Here we developed visual/motor spiking neuronal network models and trained them to play a virtual racket-ball game using several reinforcement learning algorithms inspired by the dopaminergic reward system. We systematically investigated how different architectures and circuit-motifs (feed-forward, recurrent, feedback) contributed to learning and performance. We also developed a new biologically-inspired learning rule that significantly enhanced performance, while reducing training time. Our models included visual areas encoding game inputs and relaying the information to motor areas, which used this information to learn to move the racket to hit the ball. Neurons in the early visual area relayed information encoding object location and motion direction across the network. Neuronal association areas encoded spatial relationships between objects in the visual scene. Motor populations received inputs from visual and association areas representing the dorsal pathway. Two populations of motor neurons generated commands to move the racket up or down. Model-generated actions updated the environment and triggered reward or punishment signals that adjusted synaptic weights so that the models could learn which actions led to reward. Here we demonstrate that our biologically-plausible learning rules were effective in training spiking neuronal network models to solve problems in dynamic environments. We used our models to dissect the circuit architectures and learning rules most effective for learning. Our model shows that learning mechanisms involving different neural circuits produce similar performance in sensory-motor tasks. In biological networks, all learning mechanisms may complement one another, accelerating the learning capabilities of animals. Furthermore, this also highlights the resilience and redundancy in biological systems.

Plasticity impairment exposes CA3 vulnerability in a hippocampal network model of mild traumatic brain injury.

Proper function of the hippocampus is critical for executing cognitive tasks such as learning and memory. Traumatic brain injury (TBI) and other neurological disorders are commonly associated with cognitive deficits and hippocampal dysfunction. Although there are many existing models of individual subregions of the hippocampus, few models attempt to integrate the primary areas into one system. In this work, we developed a computational model of the hippocampus, including the dentate gyrus, CA3, and CA1. The subregions are represented as an interconnected neuronal network, incorporating well-characterized ex vivo slice electrophysiology into the functional neuron models and well-documented anatomical connections into the network structure. In addition, since plasticity is foundational to the role of the hippocampus in learning and memory as well as necessary for studying adaptation to injury, we implemented spike-timing-dependent plasticity among the synaptic connections. Our model mimics key features of hippocampal activity, including signal frequencies in the theta and gamma bands and phase-amplitude coupling in area CA1. We also studied the effects of spike-timing-dependent plasticity impairment, a potential consequence of TBI, in our model and found that impairment decreases broadband power in CA3 and CA1 and reduces phase coherence between these two subregions, yet phase-amplitude coupling in CA1 remains intact. Altogether, our work demonstrates characteristic hippocampal activity with a scaled network model of spiking neurons and reveals the sensitive balance of plasticity mechanisms in the circuit through one manifestation of mild traumatic injury.

Reconstruction of a Fully Paralleled Auditory Spiking Neural Network and FPGA Implementation.

This paper presents a field-programmable gate array (FPGA) implementation of an auditory system, which is biologically inspired and has the advantages of robustness and anti-noise ability. We propose an FPGA implementation of an eleven-channel hierarchical spiking neuron network (SNN) model, which has a sparsely connected architecture with low power consumption. According to the mechanism of the auditory pathway in human brain, spiking trains generated by the cochlea are analyzed in the hierarchical SNN, and the specific word can be identified by a Bayesian classifier. Modified leaky integrate-and-fire (LIF) model is used to realize the hierarchical SNN, which achieves both high efficiency and low hardware consumption. The hierarchical SNN implemented on FPGA enables the auditory system to be operated at high speed and can be interfaced and applied with external machines and sensors. A set of speech from different speakers mixed with noise are used as input to test the performance our system, and the experimental results show that the system can classify words in a biologically plausible way with the presence of noise. The method of our system is flexible and the system can be modified into desirable scale. These confirm that the proposed biologically plausible auditory system provides a better method for on-chip speech recognition. Compare to the state-of-the-art, our auditory system achieves a higher speed with a maximum frequency of 65.03 MHz and a lower energy consumption of 276.83 μJ for a single operation. It can be applied in the field of brain-computer interface and intelligent robots.

Mean-Field Models for EEG/MEG: From Oscillations to Waves

Neural mass models have been used since the 1970s to model the coarse-grained activity of large populations of neurons. They have proven especially fruitful for understanding brain rhythms. However, although motivated by neurobiological considerations they are phenomenological in nature, and cannot hope to recreate some of the rich repertoire of responses seen in real neuronal tissue. Here we consider a simple spiking neuron network model that has recently been shown to admit an exact mean-field description for both synaptic and gap-junction interactions. The mean-field model takes a similar form to a standard neural mass model, with an additional dynamical equation to describe the evolution of within-population synchrony. As well as reviewing the origins of this next generation mass model we discuss its extension to describe an idealised spatially extended planar cortex. To emphasise the usefulness of this model for EEG/MEG modelling we show how it can be used to uncover the role of local gap-junction coupling in shaping large scale synaptic waves.

Flexible Frequency Switching in Adult Mouse Visual Cortex Is Mediated by Competition Between Parvalbumin and Somatostatin Expressing Interneurons.

Neuronal networks in rodent primary visual cortex (V1) can generate oscillations in different frequency bands depending on the network state and the level of visual stimulation. High-frequency gamma rhythms, for example, dominate the network's spontaneous activity in adult mice but are attenuated upon visual stimulation, during which the network switches to the beta band instead. The spontaneous local field potential (LFP) of juvenile mouse V1, however, mainly contains beta rhythms and presenting a stimulus does not elicit drastic changes in network oscillations. We study, in a spiking neuron network model, the mechanism in adult mice allowing for flexible switches between multiple frequency bands and contrast this to the network structure in juvenile mice that lack this flexibility. The model comprises excitatory pyramidal cells (PCs) and two types of interneurons: the parvalbumin-expressing (PV) and the somatostatinexpressing (SOM) interneuron. In accordance with experimental findings, the pyramidal-PV and pyramidal-SOM cell subnetworks are associated with gamma and beta oscillations, respectively. In our model, they are both generated via a pyramidal-interneuron gamma (PING) mechanism, wherein the PCs drive the oscillations. Furthermore, we demonstrate that large but not small visual stimulation activates SOM cells, which shift the frequency of resting-state gamma oscillations produced by the pyramidal-PV cell subnetwork so that beta rhythms emerge. Finally, we show that this behavior is obtained for only a subset of PV and SOM interneuron projection strengths, indicating that their influence on the PCs should be balanced so that they can compete for oscillatory control of the PCs. In sum, we propose a mechanism by which visual beta rhythms can emerge from spontaneous gamma oscillations in a network model of the mouse V1; for this mechanism to reproduce V1 dynamics in adult mice, balance between the effective strengths of PV and SOM cells is required.

Learning compositional sequences with multiple time scales through a hierarchical network of spiking neurons.

Sequential behaviour is often compositional and organised across multiple time scales: a set of individual elements developing on short time scales (motifs) are combined to form longer functional sequences (syntax). Such organisation leads to a natural hierarchy that can be used advantageously for learning, since the motifs and the syntax can be acquired independently. Despite mounting experimental evidence for hierarchical structures in neuroscience, models for temporal learning based on neuronal networks have mostly focused on serial methods. Here, we introduce a network model of spiking neurons with a hierarchical organisation aimed at sequence learning on multiple time scales. Using biophysically motivated neuron dynamics and local plasticity rules, the model can learn motifs and syntax independently. Furthermore, the model can relearn sequences efficiently and store multiple sequences. Compared to serial learning, the hierarchical model displays faster learning, more flexible relearning, increased capacity, and higher robustness to perturbations. The hierarchical model redistributes the variability: it achieves high motif fidelity at the cost of higher variability in the between-motif timings.

Optimal Interplay between Synaptic Strengths and Network Structure Enhances Activity Fluctuations and Information Propagation in Hierarchical Modular Networks.

In network models of spiking neurons, the joint impact of network structure and synaptic parameters on activity propagation is still an open problem. Here, we use an information-theoretical approach to investigate activity propagation in spiking networks with a hierarchical modular topology. We observe that optimized pairwise information propagation emerges due to the increase of either (i) the global synaptic strength parameter or (ii) the number of modules in the network, while the network size remains constant. At the population level, information propagation of activity among adjacent modules is enhanced as the number of modules increases until a maximum value is reached and then decreases, showing that there is an optimal interplay between synaptic strength and modularity for population information flow. This is in contrast to information propagation evaluated among pairs of neurons, which attains maximum value at the maximum values of these two parameter ranges. By examining the network behavior under the increase of synaptic strength and the number of modules, we find that these increases are associated with two different effects: (i) the increase of autocorrelations among individual neurons and (ii) the increase of cross-correlations among pairs of neurons. The second effect is associated with better information propagation in the network. Our results suggest roles that link topological features and synaptic strength levels to the transmission of information in cortical networks.

Cholinergic Switch between Two Types of Slow Waves in Cerebral Cortex.

Sleep slow waves are known to participate in memory consolidation, yet slow waves occurring under anesthesia present no positive effects on memory. Here, we shed light onto this paradox, based on a combination of extracellular recordings in vivo, in vitro, and computational models. We find two types of slow waves, based on analyzing the temporal patterns of successive slow-wave events. The first type is consistently observed in natural slow-wave sleep, while the second is shown to be ubiquitous under anesthesia. Network models of spiking neurons predict that the two slow wave types emerge due to a different gain on inhibitory versus excitatory cells and that different levels of spike-frequency adaptation in excitatory cells can account for dynamical distinctions between the two types. This prediction was tested in vitro by varying adaptation strength using an agonist of acetylcholine receptors, which demonstrated a neuromodulatory switch between the two types of slow waves. Finally, we show that the first type of slow-wave dynamics is more sensitive to external stimuli, which can explain how slow waves in sleep and anesthesia differentially affect memory consolidation, as well as provide a link between slow-wave dynamics and memory diseases.

A Thalamo-Cortex Microcircuit Model of Beta Oscillations in the Parkinsonian Motor Cortex.

Exaggerated beta oscillations (~13-30 Hz) observed in the cortical areas of the brain is one of the characteristics of disrupted information flow in the primary motor cortex in Parkinson's disease (PD). However, the mechanism underlying the generation of these enhanced beta rhythms remains unclear. The thalamo-cortex microcircuit (TCM) contains reciprocal synaptic connections that generate low frequency oscillations in the microcircuit in healthy conditions. Recent studies suggest that alterations in synaptic connections both within and between the cortex and thalamus play a critical role in the generation of pathological beta rhythms in PD. In this study, we examine this hypothesis in a spiking neuronal network model of the TCM. The model is compared and validated against neural firing patterns recorded in rodent models of PD from the literature.

A Computational Model of Loss of Dopaminergic Cells in Parkinson's Disease Due to Glutamate-Induced Excitotoxicity.

Parkinson's disease (PD) is a neurodegenerative disease associated with progressive and inexorable loss of dopaminergic cells in Substantia Nigra pars compacta (SNc). Although many mechanisms have been suggested, a decisive root cause of this cell loss is unknown. A couple of the proposed mechanisms, however, show potential for the development of a novel line of PD therapeutics. One of these mechanisms is the peculiar metabolic vulnerability of SNc cells compared to other dopaminergic clusters; the other is the SubThalamic Nucleus (STN)-induced excitotoxicity in SNc. To investigate the latter hypothesis computationally, we developed a spiking neuron network-model of SNc-STN-GPe system. In the model, prolonged stimulation of SNc cells by an overactive STN leads to an increase in ‘stress' variable; when the stress in a SNc neuron exceeds a stress threshold, the neuron dies. The model shows that the interaction between SNc and STN involves a positive-feedback due to which, an initial loss of SNc cells that crosses a threshold causes a runaway-effect, leading to an inexorable loss of SNc cells, strongly resembling the process of neurodegeneration. The model further suggests a link between the two aforementioned mechanisms of SNc cell loss. Our simulation results show that the excitotoxic cause of SNc cell loss might initiate by weak-excitotoxicity mediated by energy deficit, followed by strong-excitotoxicity, mediated by a disinhibited STN. A variety of conventional therapies were simulated to test their efficacy in slowing down SNc cell loss. Among them, glutamate inhibition, dopamine restoration, subthalamotomy and deep brain stimulation showed superior neuroprotective-effects in the proposed model.

An Oscillatory Neural Autoencoder Based on Frequency Modulation and Multiplexing.

Oscillatory phenomena are ubiquitous in the brain. Although there are oscillator-based models of brain dynamics, their universal computational properties have not been explored much unlike in the case of rate-coded and spiking neuron network models. Use of oscillator-based models is often limited to special phenomena like locomotor rhythms and oscillatory attractor-based memories. If neuronal ensembles are taken to be the basic functional units of brain dynamics, it is desirable to develop oscillator-based models that can explain a wide variety of neural phenomena. Autoencoders are a special type of feed forward networks that have been used for construction of large-scale deep networks. Although autoencoders based on rate-coded and spiking neuron networks have been proposed, there are no autoencoders based on oscillators. We propose here an oscillatory neural network model that performs the function of an autoencoder. The model is a hybrid of rate-coded neurons and neural oscillators. Input signals modulate the frequency of the neural encoder oscillators. These signals are then multiplexed using a network of rate-code neurons that has afferent Hebbian and lateral anti-Hebbian connectivity, termed as Lateral Anti Hebbian Network (LAHN). Finally the LAHN output is de-multiplexed using an output neural layer which is a combination of adaptive Hopf and Kuramoto oscillators for the signal reconstruction. The Kuramoto-Hopf combination performing demodulation is a novel way of describing a neural phase-locked loop. The proposed model is tested using both synthetic signals and real world EEG signals. The proposed model arises out of the general motivation to construct biologically inspired, oscillatory versions of some of the standard neural network models, and presents itself as an autoencoder network based on oscillatory neurons applicable to time series signals. As a demonstration, the model is applied to compression of EEG signals.

Extraction of Inter-Aural Time Differences Using a Spiking Neuron Network Model of the Medial Superior Olive.

The mammalian auditory system is able to extract temporal and spectral features from sound signals at the two ears. One important cue for localization of low-frequency sound sources in the horizontal plane are inter-aural time differences (ITDs) which are first analyzed in the medial superior olive (MSO) in the brainstem. Neural recordings of ITD tuning curves at various stages along the auditory pathway suggest that ITDs in the mammalian brainstem are not represented in form of a Jeffress-type place code. An alternative is the hemispheric opponent-channel code, according to which ITDs are encoded as the difference in the responses of the MSO nuclei in the two hemispheres. In this study, we present a physiologically-plausible, spiking neuron network model of the mammalian MSO circuit and apply two different methods of extracting ITDs from arbitrary sound signals. The network model is driven by a functional model of the auditory periphery and physiological models of the cochlear nucleus and the MSO. Using a linear opponent-channel decoder, we show that the network is able to detect changes in ITD with a precision down to 10 μs and that the sensitivity of the decoder depends on the slope of the ITD-rate functions. A second approach uses an artificial neuronal network to predict ITDs directly from the spiking output of the MSO and ANF model. Using this predictor, we show that the MSO-network is able to reliably encode static and time-dependent ITDs over a large frequency range, also for complex signals like speech.

A model of the neuronal processing of interaural time disparities

The auditory system of humans and other mammals is able to use interaural time and intensity differences as well as spectral cues to localize sound sources. One important cue for localizing low-frequency sound sources in the horizontal plane are interaural time differences (ITDs), which are first analyzed in the medial superior olive (MSO) in the brainstem. Results from electrophysiological and psychoacoustic studies suggest ITD encoding in the relative activities of neuronal populations in the two brain hemispheres. This contribution first explores the neuronal encoding of fine-structure ITDs using a physiologically motivated spiking neuronal-network model of the mammalian MSO. Results from this model confirm robust ITD encoding in the relative activity of the MSOs in both hemispheres. Based on the neuronal-network simulations, a simple probabilistic model of subsequent ITD processing is proposed. This simplified model connects the neuronal responses of the two hemispheres with different hearing sensation properties, such as lateral position, spatial extent, or number of spatially separable sensations. First predictions from the model are discussed with regard to the results of a series of psychoacoustic experiments on the lateralization of pure-tone impulses presented via headphones.

Constructing Neuronal Network Models in Massively Parallel Environments.

Recent advances in the development of data structures to represent spiking neuron network models enable us to exploit the complete memory of petascale computers for a single brain-scale network simulation. In this work, we investigate how well we can exploit the computing power of such supercomputers for the creation of neuronal networks. Using an established benchmark, we divide the runtime of simulation code into the phase of network construction and the phase during which the dynamical state is advanced in time. We find that on multi-core compute nodes network creation scales well with process-parallel code but exhibits a prohibitively large memory consumption. Thread-parallel network creation, in contrast, exhibits speedup only up to a small number of threads but has little overhead in terms of memory. We further observe that the algorithms creating instances of model neurons and their connections scale well for networks of ten thousand neurons, but do not show the same speedup for networks of millions of neurons. Our work uncovers that the lack of scaling of thread-parallel network creation is due to inadequate memory allocation strategies and demonstrates that thread-optimized memory allocators recover excellent scaling. An analysis of the loop order used for network construction reveals that more complex tests on the locality of operations significantly improve scaling and reduce runtime by allowing construction algorithms to step through large networks more efficiently than in existing code. The combination of these techniques increases performance by an order of magnitude and harnesses the increasingly parallel compute power of the compute nodes in high-performance clusters and supercomputers.

Macroscopic neural mass model constructed from a current-based network model of spiking neurons.

Neural mass models (NMMs) are efficient frameworks for describing macroscopic cortical dynamics including electroencephalogram and magnetoencephalogram signals. Originally, these models were formulated on an empirical basis of synaptic dynamics with relatively long time constants. By clarifying the relations between NMMs and the dynamics of microscopic structures such as neurons and synapses, we can better understand cortical and neural mechanisms from a multi-scale perspective. In a previous study, the NMMs were analytically derived by averaging the equations of synaptic dynamics over the neurons in the population and further averaging the equations of the membrane-potential dynamics. However, the averaging of synaptic current assumes that the neuron membrane potentials are nearly time invariant and that they remain at sub-threshold levels to retain the conductance-based model. This approximation limits the NMM to the non-firing state. In the present study, we newly propose a derivation of a NMM by alternatively approximating the synaptic current which is assumed to be independent of the membrane potential, thus adopting a current-based model. Our proposed model releases the constraint of the nearly constant membrane potential. We confirm that the obtained model is reducible to the previous model in the non-firing situation and that it reproduces the temporal mean values and relative power spectrum densities of the average membrane potentials for the spiking neurons. It is further ensured that the existing NMM properly models the averaged dynamics over individual neurons even if they are spiking in the populations.

Fundamental Activity Constraints Lead to Specific Interpretations of the Connectome.

The continuous integration of experimental data into coherent models of the brain is an increasing challenge of modern neuroscience. Such models provide a bridge between structure and activity, and identify the mechanisms giving rise to experimental observations. Nevertheless, structurally realistic network models of spiking neurons are necessarily underconstrained even if experimental data on brain connectivity are incorporated to the best of our knowledge. Guided by physiological observations, any model must therefore explore the parameter ranges within the uncertainty of the data. Based on simulation results alone, however, the mechanisms underlying stable and physiologically realistic activity often remain obscure. We here employ a mean-field reduction of the dynamics, which allows us to include activity constraints into the process of model construction. We shape the phase space of a multi-scale network model of the vision-related areas of macaque cortex by systematically refining its connectivity. Fundamental constraints on the activity, i.e., prohibiting quiescence and requiring global stability, prove sufficient to obtain realistic layer- and area-specific activity. Only small adaptations of the structure are required, showing that the network operates close to an instability. The procedure identifies components of the network critical to its collective dynamics and creates hypotheses for structural data and future experiments. The method can be applied to networks involving any neuron model with a known gain function.