Search

Relating neural activity to behavior requires an understanding of how neural computations arise from the coordinated dynamics of distributed, recurrently connected neural populations. However, inferring the nature of recurrent dynamics from partial recordings of a neural circuit presents significant challenges. Here, we show that some of these challenges can be overcome by a fine-grained analysis of the dynamics of neural residuals, i.e. trial-by-trial variability around the mean neural population trajectory for a given task condition. Residual dynamics in macaque pre-frontal cortex (PFC) in a saccade-based perceptual decision-making task reveals recurrent dynamics that is time-dependent, but consistently stable, and suggests that pronounced rotational structure in PFC trajectories during saccades is driven by inputs from upstream areas. The properties of residual dynamics restrict the possible contributions of PFC to decision-making and saccade generation, and suggest a path towards fully characterizing distributed neural computations with large-scale neural recordings and targeted causal perturbations.

Neurons are very complicated computational devices, incorporating numerous non-linear processes, particularly in their dendrites. Biophysical models capture these processes directly by explicitly modelling physiological variables, such as ion channels, current flow, membrane capacitance, etc. However, another option for capturing the complexities of real neural computation is to use cascade models, which treat individual neurons as a cascade of linear and non-linear operations, akin to a multi-layer artificial neural network. Recent research has shown that cascade models can capture single-cell computation well, but there are still a number of sub-cellular, regenerative dendritic phenomena that they cannot capture, such as the interaction between sodium, calcium, and NMDA spikes in different compartments. Here, we propose that it is possible to capture these additional phenomena usingparallel, recurrentcascade models, wherein an individual neuron is modelled as a cascade of parallel linear and non-linear operations that can be connected recurrently, akin to a multi-layer, recurrent, artificial neural network. Given their tractable mathematical structure, we show that neuron models expressed in terms of parallel recurrent cascades can themselves be integrated into multi-layered artificial neural networks and trained to perform complex tasks. We go on to discuss potential implications and uses of these models for artificial intelligence. Overall, we argue that parallel, recurrent cascade models provide an important, unifying tool for capturing single-cell computation and exploring the algorithmic implications of physiological phenomena.

Real-world tasks require coordination of working memory, decision-making, and planning, yet these cognitive functions have disproportionately been studied as independent modular processes in the brain. Here, we propose that contingency representations, defined as mappings for how future behaviors depend on upcoming events, can unify working memory and planning computations. We designed a task capable of disambiguating distinct types of representations. In task-optimized recurrent neural networks, we investigated possible circuit mechanisms for contingency representations and found that these representations can explain neurophysiological observations from the prefrontal cortex during working memory tasks. Our experiments revealed that human behavior is consistent with contingency representations and not with traditional sensory models of working memory. Finally, we generated falsifiable predictions for neural data to identify contingency representations in neural data and to dissociate different models of working memory. Our findings characterize a neural representational strategy that can unify working memory, planning, and context-dependent decision-making.

The brain constantly processes information encoded in temporal sequences of spiking activity. This sequential activity emerges from sensory inputs as well as from the brain’s own recurrent connectivity and spans multiple dynamically changing timescales. Decoding the temporal order of spiking activity across these varying timescales is a critical function of the brain, but we do not yet understand its neural implementation. The problem is, that the passive dynamics of neural membrane potentials occur on a short millisecond timescale, whereas many cognitive tasks require the integration of information across much slower behavioral timescales. However, actively generated dendritic plateau potentials do occur on such longer timescales, and their essential role for many aspects of cognition has been firmly established by recent experiments. Here, we build on these discoveries and propose a new model of neural computation that emerges from the interaction of localized plateau potentials across a functionally compartmentalized dendritic tree. We show how this interaction offers a robust solution to the timing invariant detection and processing of sequential spike patterns in single neurons. Stochastic synaptic transmission complements the deterministic all-or-none plateau process and improves information transmission by allowing ensembles of neurons to produce graded responses to continuous combinations of features. We found that networks of such neurons can solve highly complex sequence detection tasks by breaking down long inputs into sequences of shorter, random features that can be classified reliably. These results suggest that active dendritic processes are fundamental to neural computation.

Understanding the evolution of cognition requires an understanding of the costs and benefits of neural computation. This requires analysis of neuronal circuitry in terms of information-processing efficiency, ultimately cashed out in terms of ATP expenditures relative to adaptive problem-solving abilities. Despite a preoccupation in neuroscience with the synapse as the source of stored neural information, it is clear that, along with synaptic weights and electrochemical dynamics, neurons have multiple mechanisms which store and process information, including 'wetware' (protein phosphorylation, gene transcription, and so on) and cell morphology (dendritic form). Insights into non-synaptic information-processing can be gained by examining the surprisingly complex abilities of single-celled organisms ('cellular cognition') because neurons share many of the same abilities. Cells provide the fundamental level at which information processing interfaces with gene expression, and cell-internal information-processing mechanisms are both powerful and energetically efficient. Understanding cellular computation should be a central goal of research on cognitive evolution.

Representation is typically taken to be importantly separate from its physical implementation. This is exemplified in Marr’s three-level framework, widely cited and often adopted in neuroscience. However, the separation between representation and physical implementation is not a necessary feature of information-processing systems. In particular, when it comes to analog computational systems, Marr’s representational/algorithmic level and implementational level collapse into a single level. Insofar as analog computation is a better way of understanding neural computation than other notions, Marr’s three-level framework must then be amended into a two-level framework. However, far from being a problem or limitation, this sheds lights on how to understand physical media as being representational, but without a separate, medium-independent representational level.

Inspired by more detailed modeling of biological neurons, Spiking neural networks (SNNs) have been investigated both as more biologically plausible and potentially more powerful models of neural computation, and also with the aim of extracting biological neurons' energy efficiency; the performance of such networks however has remained lacking compared to classical artificial neural networks (ANNs). Here, we demonstrate how a novel surrogate gradient combined with recurrent networks of tunable and adaptive spiking neurons yields state-of-the-art for SNNs on challenging benchmarks in the time-domain, like speech and gesture recognition. This also exceeds the performance of standard classical recurrent neural networks (RNNs) and approaches that of the best modern ANNs. As these SNNs exhibit sparse spiking, we show that they theoretically are one to three orders of magnitude more computationally efficient compared to RNNs with comparable performance. Together, this positions SNNs as an attractive solution for AI hardware implementations., Comment: 11 pages

Although bottom-up attention can improve visual performance with and without awareness to the exogenous cue, whether they are governed by a common neural computation remains unclear. Using a modified Posner paradigm with backward masking, we found that the cueing effect displayed a monotonic gradient profile (Gaussian-like), both with and without awareness, whose scope, however, was significantly wider with than without awareness. This awareness-dependent scope offered us a unique opportunity to change the relative size of the attention field to the stimulus, differentially modulating the gain of attentional selection, as proposed by the normalization model of attention. Therefore, for each human subject (male and female), the stimulus size was manipulated as their respective mean attention fields with and without awareness while stimulus contrast was varied in a spatial cueing task. By measuring the gain pattern of contrast-response functions on the spatial cueing effect derived by visible or invisible cues, we observed changes in the cueing effect consonant with changes in contrast gain for visible cues and response gain for invisible cues. Importantly, a complementary analysis confirmed that subjects' awareness-dependent attention fields can be simulated by using the normalization model of attention. Together, our findings indicate an awareness-dependent normalization framework of visual bottom-up attention, placing a necessary constraint, namely, awareness, on our understanding of the neural computations underlying visual attention.SIGNIFICANCE STATEMENTBottom-up attention is known to improve visual performance with and without awareness. We discovered that manipulating subjects' awareness can modulate their attention fields of visual bottom-up attention, which offers a unique opportunity to regulate its normalization processes. On the one hand, by measuring the gain pattern of contrast-response functions on the spatial cueing effect derived by visible or invisible cues, we observed changes in the cueing effect consonant with changes in contrast gain for visible cues and response gain for invisible cues. On the other hand, by using the normalization model of attention, subjects' awareness-dependent attention fields can be simulated successfully. Our study supports important predictions of the normalization model of visual bottom-up attention and further reveals its dependence on awareness.

Neural computation has evolved to optimize the behaviors that enable our survival. Although much previous work in neuroscience has focused on constrained task behaviors, recent advances in computer vision are fueling a trend toward the study of naturalistic behaviors. Automated tracking of fine-scale behaviors is generating rich datasets for animal models including rodents, fruit flies, zebrafish, and worms. However, extracting meaning from these large and complex data often requires sophisticated computational techniques. Here we review the latest methods and modeling approaches providing new insights into the brain from behavior. We focus on unsupervised methods for identifying stereotyped behaviors and for resolving details of the structure and dynamics of behavioral sequences.

Spiking Neural Networks (SNNs) represent a biologically inspired computation model capable of emulating neural computation in human brain and brain-like structures. The main promise is very low energy consumption. Classic Von Neumann architecture based SNN accelerators in hardware, however, often fall short of addressing demanding computation and data transfer requirements efficiently at scale. In this article, we propose a promising alternative to overcome scalability limitations, based on a network of in-memory SNN accelerators, which can reduce the energy consumption by up to 150.25= when compared to a representative ASIC solution. The significant reduction in energy comes from two key aspects of the hardware design to minimize data communication overheads: (1) each node represents an in-memory SNN accelerator based on a spintronic Computational RAM array, and (2) a novel, De Bruijn graph based architecture establishes the SNN array connectivity.

The spontaneous dynamics of the brain modulate its function from moment to moment, shaping neural computation and cognition. Functional MRI (fMRI), while classically used as a tool for spatial localization, is increasingly being used to identify the temporal dynamics of brain activity. fMRI analyses focused on the temporal domain have revealed important new information about the dynamics underlying states such as arousal, attention, and sleep. Dense temporal sampling – either by using fast fMRI acquisition, or multiple repeated scan sessions within individuals – can further enrich the information present in these studies. This review focuses on recent developments in using fMRI to identify dynamics across brain states, particularly vigilance and sleep states, and the potential for highly temporally sampled fMRI to answer these questions.

Though the temporal precision of neural computation has been studied intensively, a data-driven determination of this precision remains a fundamental challenge. Reproducible spike time patterns may be obscured on single trials by uncontrolled temporal variability in behavior and cognition, or may not even be time locked to measurable signatures in either behavior or local field potentials (LFP). To overcome these challenges, we describe a general-purpose time warping framework that reveals precise spike-time patterns in an unsupervised manner, even when spiking is decoupled from behavior or is temporally stretched across single trials. We demonstrate this method across diverse systems: cued reaching in nonhuman primates, motor sequence production in rats, and olfaction in mice. This approach flexibly uncovers diverse dynamical firing patterns, including pulsatile responses to behavioral events, LFP-aligned oscillatory spiking, and even unanticipated patterns, like 7 Hz oscillations in rat motor cortex that are not time-locked to measured behaviors or LFP.

Spike trains (STs) induced by external stimuli are complex and challenging to decode the embedded spatiotemporal information. Although various methods have been developed to characterize STs, few applications have been reported in tactile discrimination applications. In this paper, a neurocomputing method based on reference spike train (RST) is proposed to establish a neural computation scheme, on which existing ST metrics could be fed into various traditional models directly. Moreover, existing metrics in the field of statistics and vector measurement are introduced together to extract more discriminative features. With the binning technique and feature selection algorithm applied, the neural computation scheme is aimed at taking advantage of as maximal as possible information contained in tactile signals. Based on our designed artificial fingertip, the effect is validated by improving the recognition accuracy from 77.6% to 83.4% when it is applied to the discrimination of eight roughness surfaces. Furthermore, properties of RST, such as spike intervals and distributions, are evaluated and it is found that RSTs with uniform distribution perform the best for tactile discrimination.

Neurophysiology and neuroanatomy constrain the set of possible computations that can be performed in a brain circuit. While detailed data on brain microcircuits is sometimes available, cognitive modelers are seldom in a position to take these constraints into account. One reason for this is the intrinsic complexity of accounting for biological mechanisms when describing cognitive function. In this paper, we present multiple extensions to the neural engineering framework (NEF), which simplify the integration of low-level constraints such as Dale's principle and spatially constrained connectivity into high-level, functional models. We focus on a model of eyeblink conditioning in the cerebellum, and, in particular, on systematically constructing temporal representations in the recurrent granule–Golgi microcircuit. We analyze how biological constraints impact these representations and demonstrate that our overall model is capable of reproducing key properties of eyeblink conditioning. Furthermore, since our techniques facilitate variation of neurophysiological parameters, we gain insights into why certain neurophysiological parameters may be as observed in nature. While eyeblink conditioning is a somewhat primitive form of learning, we argue that the same methods apply for more cognitive models as well. We implemented our extensions to the NEF in an open-source software library named “NengoBio” and hope that this work inspires similar attempts to bridge low-level biological detail and high-level function., “Best of Papers" from the 18th International Conference on Cognitive Modeling, July 20-31, 2021, Held Online

The sensory response threshold is a fundamental biophysical property of biological systems that underlies many physiological and computational functions, and its systematic study has played a pivotal role in uncovering the principles of neural computation. Here, we show that ant colonies, which perform computational tasks at the group level, have emergent collective sensory response thresholds. Colonies respond collectively to step changes in temperature and evacuate the nest during severe perturbations. This response is characterized by a group-size dependent threshold, and the underlying dynamics are dominated by social feedback between the ants. Using a binary network model, we demonstrate that a balance between short-range excitatory and long-range inhibitory interactions can explain the emergence of the collective response threshold and its size dependency. Our findings illustrate how simple social dynamics allow insect colonies to integrate information about the external environment and their internal state to produce adaptive collective responses.

Uncertainty is a fundamental property of neural computation that becomes amplified when sensory information does not match a person’s expectations of the world. Uncertainty and hesitation are often early indicators of potential disruption, and the ability to rapidly measure uncertainty would have implications for future educational and training efforts by targeting reflective discussions about past actions, supporting in-progress corrections, and generating forecasts about future disruptions. An approach is described combining neurodynamics and machine learning to provide quantitative measures of uncertainty. Models of neurodynamic information derived from electroencephalogram (EEG) brainwaves have provided detailed neurodynamic histories of US Navy submarine navigation team members. Persistent periods (25–30 s) of neurodynamic information were seen as discrete peaks when establishing the submarine’s position and were identified as periods of uncertainty by an artificial intelligence (AI) system previously trained to recognize the frequency, magnitude, and duration of different patterns of uncertainty in healthcare and student teams. Transition matrices of neural network states closely predicted the future uncertainty of the navigation team during the three minutes prior to a grounding event. These studies suggest that the dynamics of uncertainty may have common characteristics across teams and tasks and that forecasts of their short-term evolution can be estimated.

In systems neuroscience, advances in simultaneous recording technology have helped reveal the population dynamics that underlie the complex neural correlates of animal behavior and cognitive processes. To investigate these correlates, neural interactions are typically abstracted from spike trains of pairs of neurons accumulated over the course of many trials. However, the resultant averaged values do not lead to understanding of neural computation in which the responses of populations are highly variable even under identical external conditions. Accordingly, neural interactions within the population also show strong fluctuations. In the present study, we introduce an analysis method reflecting the temporal variation of neural interactions, in which cross-correlograms on rate estimates are applied via a latent dynamical systems model. Using this method, we were able to predict time-varying neural interactions within a single trial. In addition, the pairwise connections estimated in our analysis increased along behavioral epochs among neurons categorized within similar functional groups. Thus, our analysis method revealed that neurons in the same groups communicate more as the population gets involved in the assigned task. We also showed that the characteristics of neural interaction from our model differ from the results of a typical model employing cross-correlation coefficients. This suggests that our model can extract nonoverlapping information about network topology, unlike the typical model.

Nonlinear interactions in the dendritic tree play a key role in neural computation. Nevertheless, modeling frameworks aimed at the construction of large-scale, functional spiking neural networks, such as the Neural Engineering Framework, tend to assume a linear superposition of postsynaptic currents. In this letter, we present a series of extensions to the Neural Engineering Framework that facilitate the construction of networks incorporating Dale's principle and nonlinear conductance-based synapses. We apply these extensions to a two-compartment LIF neuron that can be seen as a simple model of passive dendritic computation. We show that it is possible to incorporate neuron models with input-dependent nonlinearities into the Neural Engineering Framework without compromising high-level function and that nonlinear postsynaptic currents can be systematically exploited to compute a wide variety of multivariate, band-limited functions, including the Euclidean norm, controlled shunting, and nonnegative multiplication. By avoiding an additional source of spike noise, the function approximation accuracy of a single layer of two-compartment LIF neurons is on a par with or even surpasses that of two-layer spiking neural networks up to a certain target function bandwidth.

Aberrations in functional connectivity in children with developmental dyslexia have been found in electroencephalographic studies using graph analysis. How training with visual tasks can modify the functional semantic network in developmental dyslexia remains unclear. We investigate local and global topological properties of functional networks in multiple EEG frequency ranges based on a small-world propensity method in controls, pre- and post-training dyslexic children during visual word/pseudoword processing. Results indicated that the EEG network topology in dyslexics before the training was more integrated than controls, and after training - more segregated and similar to that of the controls in the theta (θ: 4-8), alpha (α: 8-13), beta (β1: 13-20; β2: 20-30), and gamma (γ1: 30-48; γ2: 52-70 Hz) bands for three graph measures. The pre-training dyslexics exhibited a reduced strength and betweenness centrality of the left anterior temporal and parietal regions in the θ, α, β1 and γ1-frequency bands, compared to the controls. The simultaneous appearance of hubs in the left hemisphere (or both hemispheres) at temporal and parietal (α-word/γ-pseudoword discrimination), temporal and middle frontal cortex (θ, α-word), parietal and middle frontal cortex (β1-word), parietal and occipitotemporal cortices (θ-pseudoword), identified in the EEG-based functional networks of normally developing children were not present in the networks of dyslexics. The hub distribution for dyslexics in the θ, α, and β1 bands became similar to that of the controls. The topological organization of functional networks and the less efficient network configuration (long characteristic path length) in dyslexics compared to the more optimal global organization in the controls was studied for the first time after remediation training.

The brain can learn new tasks without forgetting old ones. This memory retention is closely associated with the long-term stability of synaptic strength. To understand the capacity of pyramidal neurons to preserve memory under different tasks, we established a plasticity model based on the postsynaptic membrane energy state, in which the change in synaptic strength depends on the difference between the energy state after stimulation and the resting energy state. If the post-stimulation energy state is higher than the resting energy state, then synaptic depression occurs. On the contrary, the synapse is strengthened. Our model unifies homo- and heterosynaptic plasticity and can reproduce synaptic plasticity observed in multiple experiments, such as spike-timing-dependent plasticity, and cooperative plasticity with few and common parameters. Based on the proposed plasticity model, we conducted a simulation study on how the activation patterns of dendritic branches by different tasks affect the synaptic connection strength of pyramidal neurons. We further investigate the formation mechanism by which different tasks activate different dendritic branches. Simulation results show that compare to the classic plasticity model, the plasticity model we proposed can achieve a better spatial separation of different branches activated by different tasks in pyramidal neurons, which deepens our insight into the memory retention mechanism of brains.

Motives motivate human behavior. Most behaviors are driven by more than one motive, yet it is unclear how different motives interact and how such motive combinations affect the neural computation of the behaviors they drive. To answer this question, we induced two prosocial motives simultaneously (multi-motive condition) and separately (single motive conditions). After the different motive inductions, participants performed the same choice task in which they allocated points in favor of the other person (prosocial choice) or in favor of themselves (egoistic choice). We used fMRI to assess prosocial choice-related brain responses and drift diffusion modelling to specify how motive combinations affect individual components of the choice process. Our results showed that the combination of the two motives in the multi-motive condition increased participants’ choice biases prior to the behavior itself. On the neural level, these changes in initial prosocial bias were associated with neural responses in the bilateral dorsal striatum. In contrast, the efficiency of the prosocial decision process was comparable between the multi-motive and the single-motive conditions. These findings provide insights into the computation of prosocial choices in complex motivational states, the motivational setting that drives most human behaviors.HighlightsActivating different social motives simultaneously can enhance prosocial choicesMulti-motive combinations change initial prosocial biasesDorso-striatal activation increases with larger increase of prosocial biasMulti-motive combinations modulate relative response caution

In this paper, considering the applications of interval analysis in various fields (such as artificial intelligence, neural computation, genetic algorithms, information theory or fuzzy logic), a new class of interval-valued variational control problems governed by multiple integral functionals, first-order PDE and inequality constraints is studied. More precisely, efficiency conditions for the considered uncertain variational control problem are formulated and proved. The sufficiency of Karush–Kuhn–Tucker conditions is established under some invexity and $$(\rho, b)$$ -quasiinvexity assumptions of the involved functionals. In addition, the paper is completed with illustrative applications (describing the controlled behavior of an artificial neural system) and the corresponding algorithm.

Understanding the mechanisms of neural computation and learning will require knowledge of the underlying circuitry. Because it is difficult to directly measure the wiring diagrams of neural circuits, there has long been an interest in estimating them algorithmically from multicell activity recordings. We show that even sophisticated methods, applied to unlimited data from every cell in the circuit, are biased toward inferring connections between unconnected but highly correlated neurons. This failure to ‘explain away’ connections occurs when there is a mismatch between the true network dynamics and the model used for inference, which is inevitable when modeling the real world. Thus, causal inference suffers when variables are highly correlated, and activity-based estimates of connectivity should be treated with special caution in strongly connected networks. Finally, performing inference on the activity of circuits pushed far out of equilibrium by a simple low-dimensional suppressive drive might ameliorate inference bias. The authors demonstrate that strongly recurrent circuits inferred from neural activity, even with unlimited data from every neuron, are biased. Synapses are inferred between unconnected but correlated neurons. Inference based on non-equilibrium activity may help remedy this.

Neural computation is associated with the emergence, reconfiguration, and dissolution of cell assemblies in the context of varying oscillatory states. Here, we describe the complex spatiotemporal dynamics of cell assemblies through temporal network formalism. We use a sliding window approach to extract sequences of networks of information sharing among single units in hippocampus and entorhinal cortex during anesthesia and study how global and node-wise functional connectivity properties evolve through time and as a function of changing global brain state (theta vs. slow-wave oscillations). First, we find that information sharing networks display, at any time, a core-periphery structure in which an integrated core of more tightly functionally interconnected units links to more loosely connected network leaves. However the units participating to the core or to the periphery substantially change across time windows, with units entering and leaving the core in a smooth way. Second, we find that discrete network states can be defined on top of this continuously ongoing liquid core-periphery reorganization. Switching between network states results in a more abrupt modification of the units belonging to the core and is only loosely linked to transitions between global oscillatory states. Third, we characterize different styles of temporal connectivity that cells can exhibit within each state of the sharing network. While inhibitory cells tend to be central, we show that, otherwise, anatomical localization only poorly influences the patterns of temporal connectivity of the different cells. Furthermore, cells can change temporal connectivity style when the network changes state. Altogether, these findings reveal that the sharing of information mediated by the intrinsic dynamics of hippocampal and entorhinal cortex cell assemblies have a rich spatiotemporal structure, which could not have been identified by more conventional time- or state-averaged analyses of functional connectivity., Author Summary It is generally thought that computations performed by local brain circuits rely on complex neural processes, associated with the flexible waxing and waning of cell assemblies, that is, an ensemble of cells firing in tight synchrony. Although cell assembly formation is inherently and unavoidably dynamical, it is still common to find studies in which essentially “static” approaches are used to characterize this process. In the present study, we adopt instead a temporal network approach. Avoiding usual time-averaging procedures, we reveal that hub neurons are not hardwired but that cells vary smoothly their degree of integration within the assembly core. Furthermore, our temporal network framework enables the definition of alternative possible styles of “hubness.” Some cells may share information with a multitude of other units but only in an intermittent manner, as “activists” in a flash mob. In contrast, some other cells may share information in a steadier manner, as resolute “lobbyists.” Finally, by avoiding averages over preimposed states, we show that within each global oscillatory state rich switching dynamics can take place between a repertoire of many available network states. We thus show that the temporal network framework provides a natural and effective language to rigorously describe the rich spatiotemporal patterns of information sharing instantiated by cell assembly evolution.

Spiking neural networks (SNNs) are believed to be a powerful neural computation framework inspired by the vivo neurons. As a class of recurrent SNNs, liquid state machines (LSMs) are biologically more plausible models imitating the architecture and functions of the human brain for information processing. However, few LSM models can outperform conventional analogue neural networks for solving real-world classification or regression problems, which can mainly be attributed to the sensitivity of the training performance to the architecture of the reservoir and the parameters in the spiking neuron models. Most recently, many algorithms have been proposed for automated machine learning that aims to automatically design the architecture and parameters of deep neural networks without much human intervention. Although automated machine learning and neural architecture search have been extremely successful in conventional neural networks, little research on search for an optimal architecture and hyperparameters of LSMs has been reported. This work proposes on a surrogate-assisted evolutionary search method for optimization of the hyperparameters and neural architecture of the reservoir of LSMs using the covariance matrix adaptation evolution strategy (CMA-ES). For reducing the search space, the architecture of the LSM is encoded by a connectivity probability together with the hyperparameters in the spiking neuron models. To enhance the computational efficiency, a Gaussian process is adopted as the surrogate to assist the CMA-ES. The proposed GP-assisted CMA-ES is compared with the canonical CMA-ES and a Bayesian optimization algorithm on two popular datasets including image and action recognition. Our results confirm that the proposed algorithm is efficient and effective in optimizing the parameters and architecture of LSMs.

Toward the long-standing dream of artificial intelligence, two successful solution paths have been paved: 1) neuromorphic computing and 2) deep learning. Recently, they tend to interact for simultaneously achieving biological plausibility and powerful accuracy. However, models from these two domains have to run on distinct substrates, i.e., neuromorphic platforms and deep learning accelerators, respectively. This architectural incompatibility greatly compromises the modeling flexibility and hinders promising interdisciplinary research. To address this issue, we build a unified model description framework and a unified processing architecture (Tianjic), which covers the full stack from software to hardware. By implementing a set of integration and transformation operations, Tianjic is able to support spiking neural networks, biological dynamic neural networks, multilayered perceptron, convolutional neural networks, recurrent neural networks, and so on. A compatible routing infrastructure enables homogeneous and heterogeneous scalability on a decentralized many-core network. Several optimization methods are incorporated, such as resource and data sharing, near-memory processing, compute/access skipping, and intra-/inter-core pipeline, to improve performance and efficiency. We further design streaming mapping schemes for efficient network deployment with a flexible tradeoff between execution throughput and resource overhead. A 28-nm prototype chip is fabricated with >610-GB/s internal memory bandwidth. A variety of benchmarks are evaluated and compared with GPUs and several existing specialized platforms. In summary, the fully unfolded mapping can achieve significantly higher throughput and power efficiency; the semi-folded mapping can save 30x resources while still presenting comparable performance on average. Finally, two hybrid-paradigm examples, a multimodal unmanned bicycle and a hybrid neural network, are demonstrated to show the potential of our unified architecture. This article paves a new way to explore neural computing.

Spiking neural P systems (SN P systems) refer to a type of distributed parallel neural computation model in the membrane computing framework. In this work, a new type of SN P system is examined, i.e., spiking neural P systems with polarizations (PSN P systems), by taking inspiration from the polarized cell membrane of a neuron. For the PSN P system, the firing condition of rules is the neuron-associated polarization. This work focuses on examining the computation power of the PSN P systems adopting the sequential mode induced by the minimum spike number, where at a computation step, one (resp., all) of neurons that hold the minimum spike number within the neurons which can fire will fire, i.e., the min-sequentiality strategy (resp., min-pseudo-sequentiality strategy). We prove that PSN P systems adopting the min-sequentiality strategy or min-pseudo-sequentiality strategy are Turing universal as the number generating devices. These results indicate the computation power of PSN P systems is robust regarding their strategies of sequentiality.

Many animals use an internal sense of direction to guide their movements through the world. Neurons selective to head direction are thought to support this directional sense and have been found in a diverse range of species, from insects to primates, highlighting their evolutionary importance. Across species, most head-direction networks share four key properties: a unique representation of direction at all times, persistent activity in the absence of movement, integration of angular velocity to update the representation, and the use of directional cues to correct drift. The dynamics of theorized network structures called ring attractors elegantly account for these properties, but their relationship to brain circuits is unclear. Here, we review experiments in rodents and flies that offer insights into potential neural implementations of ring attractor networks. We suggest that a theory-guided search across model systems for biological mechanisms that enable such dynamics would uncover general principles underlying head-direction circuit function.

In sensorimotor control, sensory feedback integrates with forward models to alleviate the impacts of sensory noise and delay on state estimation. The sensorimotor integration is subject to Bayesian inference and has been formulated by the Kalman filter in computational neuroscience. However, the Kalman filter, as an artificial optimal estimator to address the abstract characteristics of spatial perception, is inadequate to present the neural computation in the cerebellum. Besides, the nonlinear neuromuscular dynamics with tightly coupled state variables also substantially impedes the implementation of Kalman filter in realistic sensorimotor systems. Here we address the sensorimotor state estimate by using the particle filter, a nonlinear Bayesian estimator that can be implemented in arbitrary dynamic systems with the neurocomputational compatibility. Particle filtering is explicitly implemented in a biophysically realistic sensorimotor model of an upper limb integrating Hill-type muscles, tendons, skeleton, and primary afferents. By involving the command noises, the constructed neuromusculoskeletal model qualitatively represents the experimental variability in center-out reaching movements. Despite the initial estimation uncertainty and sensorimotor noises, the particle filter is able to approximate the actual states in forward-reaching movements. Furthermore, the simulated hand-position estimate is consistent with the experimental results, in the presence of forward model errors, neural noises, and sensory delays. The particle filter is demonstrated to effectively implement the Bayesian state estimation in biophysically realistic sensorimotor systems and provide better compatibility with neuronal computation than the Kalman filter.

Since the first quantum neural network based on quantum computing was proposed by famous scholar Kak, much attention has been taken focus on designing new quantum neural network models. In this paper, a novel efficient quantum feed-forward neural network based on quantum computing is established, which adopts genetic algorithm to improve the traditional back propagation algorithm in parameters learning process. We clearly show the mathematical process of the new proposed quantum network model and improved algorithm. The experimental results of MATLAB simulations show that the new network model which makes the best use of fast quantum neural computation does a better job in function approximation and prediction of educational short video’s spreading capacity than traditional back propagation neural network, and the improved algorithm is more efficient than common back propagation algorithm in the proposed quantum network model. Our model can be widely used in weather prediction, handwriting recognition, speech recognition, and other aspects.

1.AbstractAt ultra-high field, fMRI voxels can span the sub-millimeter range, allowing the recording of blood oxygenation level dependent (BOLD) responses at the level of fundamental units of neural computation, such as cortical columns and layers. This sub-millimeter resolution, however, is only nominal in nature as a number of factors limit the spatial acuity of functional voxels. Multivoxel Pattern Analysis (MVPA) may provide a means to detect information at finer spatial scales that may otherwise not be visible at the single voxel level due to limitations in sensitivity and specificity. Here, we evaluate the spatial scale of stimuli specific BOLD responses in multivoxel patterns exploited by linear Support Vector Machine, Linear Discriminant Analysis and Naïve Bayesian classifiers across cortical depths in V1. To this end, we artificially misaligned the testing relative to the training portion of the data in increasing spatial steps, then investigated the breakdown of the classifiers’ performances. A one voxel shift led to a significant decrease in decoding accuracy (p

Focusing on biolinguistic considerations, Merge is a formal operation proposed by theoretical linguistics and linked to specific principles of neural computation. In this sense, Merge can be viewed as a natural operation of the brain. Merge is commonly claimed to be a digital operation. To a first approximation, digital computation is the processing of strings of digits according to general rules defined over these digits. However, it seems that neural processes are not digital computations. These conflictive claims, e.g. the digital characterization of Merge and the non-digital characterization of the brain, leads to the following scenario: either Merge is an operation that is not realized in the brain or Merge is realized in the brain but not digitally. The purpose of this paper is to evaluate the problems of each thesis.

Efficient data collection is an important goal in cognitive neuroimaging studies because of the high cost of data acquisition. One method of improving efficiency is to maximize the informativeness of the data collected on each trial. We propose an Adaptive Design Optimization (Cavagnaro et al. Neural Computation 22, 887–905 2010; Myung et al. Journal of Mathematical Psychology 57, 53–67 2013) procedure to optimize the sequencing of stimuli for model-based functional neuroimaging studies. Our method uses the Joint Modeling Framework (Turner et al. NeuroImage 72, 193–206 2013, 2019) to maximize the information learned about how the brain produces a behavior by integrating over neural and behavioral data simultaneously. We validate our method in simulation and real-world experiments by showing how Adaptive Design Optimization proposes the optimal stimulus sequence to reduce uncertainty and improve accuracy from a Bayesian perspective.

Neural computation relies on the integration of synaptic inputs across a neuron’s dendritic arbour. However, it is far from understood how different cell types tune this process to establish cell-type specific computations. Here, using two-photon imaging of dendritic Ca2+ signals, electrical recordings of somatic voltage and biophysical modelling, we demonstrate that four morphologically distinct types of mouse retinal ganglion cells with overlapping excitatory synaptic input (transient Off alpha, transient Off mini, sustained Off, and F-mini Off) exhibit type-specific dendritic integration profiles: in contrast to the other types, dendrites of transient Off alpha cells were spatially independent, with little receptive field overlap. The temporal correlation of dendritic signals varied also extensively, with the highest and lowest correlation in transient Off mini and transient Off alpha cells, respectively. We show that differences between cell types can likely be explained by differences in backpropagation efficiency, arising from the specific combinations of dendritic morphology and ion channel densities., Neurons compute by integrating synaptic inputs across their dendritic arbor. Here, the authors show that distinct cell-types of mouse retinal ganglion cells that receive similar excitatory inputs have different biophysical mechanisms of input integration to generate their unique response tuning.

The prediction of convective clouds formation is a very important problem in different areas such as agriculture, natural hazards prevention or transport-related facilities. In this paper, we evaluate the capacity of different types of evolutionary artificial neural networks to predict the formation of convective clouds, tackling the problem as a classification task. We use data from Madrid-Barajas airport, including variables and indices derived from the Madrid-Barajas airport radiosonde station. As objective variable, we use the cloud information contained in the METAR and SPECI meteorological reports from the same airport and we consider a prediction time horizon of 12 h. The performance of different types of evolutionary artificial neural networks has been discussed and analysed, including three types of basis functions (sigmoidal unit, product unit and radial basis function) and two types of models, a mono-objective evolutionary algorithm with two objective functions and a multi-objective evolutionary algorithm optimised by the two objective functions simultaneously. We show that some of the developed neuro-evolutionary models obtain high quality solutions to this problem, due to its high unbalance characteristic.

Hybrid composites are advanced composites in which the required characteristics of the material can be obtained easily by choice of suitable combinations of fibers and fillers. In this study, slidi...

According to the critical brain hypothesis, the brain is considered to operate near criticality and realize efficient neural computations. Despite the prior theoretical and empirical evidence in favor of the hypothesis, no direct link has been provided between human cognitive performance and the neural criticality. Here we provide such a key link by analyzing resting-state dynamics of functional magnetic resonance imaging (fMRI) networks at a whole-brain level. We develop a data-driven analysis method, inspired from statistical physics theory of spin systems, to map out the whole-brain neural dynamics onto a phase diagram. Using this tool, we show evidence that neural dynamics of human participants with higher fluid intelligence quotient scores are closer to a critical state, i.e., the boundary between the paramagnetic phase and the spin-glass (SG) phase. The present results are consistent with the notion of “edge-of-chaos” neural computation., Ezaki et al. develop a computational tool to analyze neural resting-state dynamics of functional magnetic resonance imaging data. Their data from adult humans suggest that the ability to think logically and find solutions improves with the brain located closer to criticality.

Optogenetic stimulation has become the method of choice for investigating neural computation in populations of neurons. Optogenetic experiments often aim to elicit a network response by stimulating specific groups of neurons. However, this is complicated by the fact that optogenetic stimulation is nonlinear, more light does not always equal to more spikes, and neurons that are not directly but indirectly stimulated could have a major impact on how networks respond to optogenetic stimulation. To clarify how optogenetic excitation of some neurons alters the network dynamics, we studied the temporal and spatial response of individual neurons and recurrent neural networks. In individual neurons, we find that neurons show a monotonic, saturating rate response to increasing light intensity and a nonmonotonic rate response to increasing pulse frequency. At the network level, we find that Gaussian light beams elicit spatial firing rate responses that are substantially broader than the stimulus profile. In summary, our analysis and our network simulation code allow us to predict the outcome of an optogenetic experiment and to assess whether the observed effects can be attributed to direct or indirect stimulation of neurons., Author Summary Optogenetic circuit manipulation has become a popular tool to manipulate the activity of neurons. During optogenetic stimulation, the firing rate of a neuron can rise because of direct light excitation or indirect activation via other neurons. To disentangle these influences and predict the effects of optogenetic excitation, we set up a spiking network model with controlled connectivity and studied its response to light stimulation. We find that the optogenetically evoked activity in a network can spread far beyond the light-stimulated area. We further found a nonmonotonic rate response of single neurons to increasing light intensities and frequencies. Our results help to interpret optogenetic experiments in vivo, and we provide computer code that can be customized to simulate 2D connectivity scenarios and explore their consequences.

Learning to select appropriate actions based on their values is fundamental to adaptive behavior. This form of learning is supported by fronto-striatal systems. The dorsal-lateral prefrontal cortex (dlPFC) and the dorsal striatum (dSTR), which are strongly interconnected, are key nodes in this circuitry. Substantial experimental evidence, including neurophysiological recordings, have shown that neurons in these structures represent key aspects of learning. The computational mechanisms that shape the neurophysiological responses, however, are not clear. To examine this, we developed a recurrent neural network (RNN) model of the dlPFC-dSTR circuit and trained it on an oculomotor sequence learning task. We compared the activity generated by the model to activity recorded from monkey dlPFC and dSTR in the same task. This network consisted of a striatal component which encoded action values, and a prefrontal component which selected appropriate actions. After training, this system was able to autonomously represent and update action values and select actions, thus being able to closely approximate the representational structure in corticostriatal recordings. We found that learning to select the correct actions drove action-sequence representations further apart in activity space, both in the model and in the neural data. The model revealed that learning proceeds by increasing the distance between sequence-specific representations. This makes it more likely that the model will select the appropriate action sequence as learning develops. Our model thus supports the hypothesis that learning in networks drives the neural representations of actions further apart, increasing the probability that the network generates correct actions as learning proceeds. Altogether, this study advances our understanding of how neural circuit dynamics are involved in neural computation, showing how dynamics in the corticostriatal system support task learning.

Waste marble dust is a solid waste generated during the cutting and polishing of marble pieces in construction sites and also in marble processing industries. This paper reports on the utilization of this waste as a filler material in particulate-filled polymer composite. Polyester-based composites are prepared with different weight proportions of waste marble dust and the dry sliding wear behavior of these composites is studied. Wear trials are conducted using a pin-on-disc test apparatus based on the Taguchi’s L25 orthogonal array as per ASTM G 99-05. The effects of different parameters on the specific wear rate of the composites are studied and an optimum combination of parameters is obtained for the least wear rate. Based on the experimental data, a prediction model using the artificial neural network is used to predict the specific wear rate of the composites at a wider range of operating parameters, within and beyond the test region. The morphologies of the worn surfaces are studied by a scanning electron microscope to ascertain the wear mechanism of the composites at different conditions. This work thus opens up a new avenue for the value added utilization of a waste like marble dust in tribological applications.

In this paper, the computational performance of a Spiking Feed-forward Neural Network (SFNN) is investigated based on a brain-inspired Intrinsic Plasticity (IP) mechanism, which is a membrane potential adaptive tuning scheme used to change the intrinsic excitability of individual neuron. This learning rule has the ability of regulating neural activity in a relative homeostatic level even if the external input of a neuron is extremely low or extremely high. The effectiveness of IP on SFNN model has been studied and evaluated through the MNIST handwritten digits classification. The training of network weights starts from a conventional artificial neural network by backpropagation and then the rate-based neurons are transformed into spiking neuron models with IP learning. Our results show that both over-activation and under-activation of neuronal response which commonly exist during the computation of neural networks can be effectively avoided. Without loss of accuracy, the calculation speed of SFNN with IP learning is extremely higher than that of the other models. Besides, when the input intensity and data noise are taken into account, both of the learning speed and accuracy of the model can be greatly improved by the application of IP learning. This biologically inspired SFNN model is simple and effective which may give insights for the optimization of neural computation.

Numerous efforts to generate "connectomes," or synaptic wiring diagrams, of large neural circuits or entire nervous systems are currently underway. These efforts promise an abundance of data to guide theoretical models of neural computation and test their predictions. However, there is not yet a standard set of tools for incorporating the connectivity constraints that these datasets provide into the models typically studied in theoretical neuroscience. This article surveys recent approaches to building models with constrained wiring diagrams and the insights they have provided. It also describes challenges and the need for new techniques to scale these approaches to ever more complex datasets.

Calcium imaging with protein-based indicators is widely used to follow neural activity in intact nervous systems. The popular GCaMP indicators are based on the calcium-binding protein calmodulin and the RS20 peptide. These sensors report neural activity at timescales much slower than electrical signaling, limited by their biophysical properties and trade-offs between sensitivity and speed. We used large-scale screening and structure-guided mutagenesis to develop and optimize several fast and sensitive GCaMP-type indicators. The resulting ‘jGCaMP8’ sensors, based on calmodulin and a fragment of endothelial nitric oxide synthase, have ultra-fast kinetics (rise times, 2 ms) and still feature the highest sensitivity for neural activity reported for any protein-based sensor. jGCaMP8 sensors will allow tracking of larger populations of neurons on timescales relevant to neural computation.

A foundational result in neural computation was proving the firing-rate model of neurons defines a universal function approximator. In this paper I prove it is possible for astrocytes to compute anything neurons can, by extending this original proof to a model of calcium waves in astrocytes. I confirm this in a series of computer simulations. The major limit for astrocytes, it turns out, is not their ability to learn solutions but the computational complexity of their network. I suggest some initial experiments that might be used to confirm these predictions.

The function of a neural circuit can be determined by the following: (1) characteristics of individual neurons composing the circuit, (2) their distinct connection structure, and (3) their neural circuit activity. However, prior research on correlations between these three factors revealed many limitations. In particular, profiling and modeling of the connectivity of complex neural circuits at the cellular level are highly challenging. To reduce the burden of the analysis, we suggest a new approach with simplification of the neural connection in an array of honeycomb patterns on 2D, using a microcontact printing technique. Through a series of guided neuronal growths in defined honeycomb patterns, a simplified neuronal circuit was achieved. Our approach allowed us to obtain the whole network connectivity at cellular resolution using a combination of stochastic multicolor labeling via viral transfection. Therefore, we were able to identify several types of hub neurons with distinct connectivity features. We also compared the structural differences between different circuits using three-node motif analysis. This new model system, iCANN, is the first experimental model of neural computation at the cellular level, providing neuronal circuit structures for the study of the relationship between anatomical structure and function of the neuronal network.

Neural coding, including encoding and decoding, is one of the key problems in neuroscience for understanding how the brain uses neural signals to relate sensory perception and motor behaviors with neural systems. However, most of the existed studies only aim at dealing with the continuous signal of neural systems, while lacking a unique feature of biological neurons, termed spike, which is the fundamental information unit for neural computation as well as a building block for brain-machine interface. Aiming at these limitations, we propose a transcoding framework to encode multi-modal sensory information into neural spikes and then reconstruct stimuli from spikes. Sensory information can be compressed into 10% in terms of neural spikes, yet re-extract 100% of information by reconstruction. Our framework can not only feasibly and accurately reconstruct dynamical visual and auditory scenes, but also rebuild the stimulus patterns from functional magnetic resonance imaging (fMRI) brain activities. More importantly, it has a superb ability of noise immunity for various types of artificial noises and background signals. The proposed framework provides efficient ways to perform multimodal feature representation and reconstruction in a high-throughput fashion, with potential usage for efficient neuromorphic computing in a noisy environment.

Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in primates, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available.Author summaryTask-driven deep neural networks have shown great potential in predicting functional responses of biological neurons. Nevertheless, they are not precise biological analogues, raising questions about how they should be interpreted. Here, we build new deep neural network models of the mouse visual cortex (MouseNet) that are biologically constrained in detail, not only in terms of the basic structure of their connectivity, but also in terms of the count and hence density of neurons within each area, and the spatial extent of their projections. Equipped with the MouseNet model, we can address key questions about mesoscale brain architecture and its role in task learning and performance.We ask, and provide a first set of answers, to: What is the performance of a mouse brain-sized – and mouse brain-structured – model on benchmark image classification tasks? How does the training of a network on this task affect the functional properties of specified layers within the biologically constrained architecture – both overall, and in comparison with recorded function of mouse neurons? We anticipate much future work on allied questions, and the development of more sophisticated models in both mouse and other species, based on the freely available MouseNet model and code which we develop and provide here.