Predictive coding with spiking neurons and feedforward gist signaling.

Predictive coding (PC) is an influential theory in neuroscience, which suggests the existence of a cortical architecture that is constantly generating and updating predictive representations of sensory inputs. Owing to its hierarchical and generative nature, PC has inspired many computational models of perception in the literature. However, the biological plausibility of existing models has not been sufficiently explored due to their use of artificial neurons that approximate neural activity with firing rates in the continuous time domain and propagate signals synchronously. Therefore, we developed a spiking neural network for predictive coding (SNN-PC), in which neurons communicate using event-driven and asynchronous spikes. Adopting the hierarchical structure and Hebbian learning algorithms from previous PC neural network models, SNN-PC introduces two novel features: (1) a fast feedforward sweep from the input to higher areas, which generates a spatially reduced and abstract representation of input (i.e., a neural code for the gist of a scene) and provides a neurobiological alternative to an arbitrary choice of priors; and (2) a separation of positive and negative error-computing neurons, which counters the biological implausibility of a bi-directional error neuron with a very high baseline firing rate. After training with the MNIST handwritten digit dataset, SNN-PC developed hierarchical internal representations and was able to reconstruct samples it had not seen during training. SNN-PC suggests biologically plausible mechanisms by which the brain may perform perceptual inference and learning in an unsupervised manner. In addition, it may be used in neuromorphic applications that can utilize its energy-efficient, event-driven, local learning, and parallel information processing nature.

A disinhibitory circuit mechanism explains a general principle of peak performance during mid-level arousal.

Perceptual decision-making is highly dependent on the momentary arousal state of the brain, which fluctuates over time on a scale of hours, minutes, and even seconds. The textbook relationship between momentary arousal and task performance is captured by an inverted U-shape, as put forward in the Yerkes-Dodson law. This law suggests optimal performance at moderate levels of arousal and impaired performance at low or high arousal levels. However, despite its popularity, the evidence for this relationship in humans is mixed at best. Here, we use pupil-indexed arousal and performance data from various perceptual decision-making tasks to provide converging evidence for the inverted U-shaped relationship between spontaneous arousal fluctuations and performance across different decision types (discrimination, detection) and sensory modalities (visual, auditory). To further understand this relationship, we built a neurobiologically plausible mechanistic model and show that it is possible to reproduce our findings by incorporating two types of interneurons that are both modulated by an arousal signal. The model architecture produces two dynamical regimes under the influence of arousal: one regime in which performance increases with arousal and another regime in which performance decreases with arousal, together forming an inverted U-shaped arousal-performance relationship. We conclude that the inverted U-shaped arousal-performance relationship is a general and robust property of sensory processing. It might be brought about by the influence of arousal on two types of interneurons that together act as a disinhibitory pathway for the neural populations that encode the available sensory evidence used for the decision.

Local minimization of prediction errors drives learning of invariant object representations in a generative network model of visual perception.

The ventral visual processing hierarchy of the cortex needs to fulfill at least two key functions: perceived objects must be mapped to high-level representations invariantly of the precise viewing conditions, and a generative model must be learned that allows, for instance, to fill in occluded information guided by visual experience. Here, we show how a multilayered predictive coding network can learn to recognize objects from the bottom up and to generate specific representations via a top-down pathway through a single learning rule: the local minimization of prediction errors. Trained on sequences of continuously transformed objects, neurons in the highest network area become tuned to object identity invariant of precise position, comparable to inferotemporal neurons in macaques. Drawing on this, the dynamic properties of invariant object representations reproduce experimentally observed hierarchies of timescales from low to high levels of the ventral processing stream. The predicted faster decorrelation of error-neuron activity compared to representation neurons is of relevance for the experimental search for neural correlates of prediction errors. Lastly, the generative capacity of the network is confirmed by reconstructing specific object images, robust to partial occlusion of the inputs. By learning invariance from temporal continuity within a generative model, the approach generalizes the predictive coding framework to dynamic inputs in a more biologically plausible way than self-supervised networks with non-local error-backpropagation. This was achieved simply by shifting the training paradigm to dynamic inputs, with little change in architecture and learning rule from static input-reconstructing Hebbian predictive coding networks.

Spontaneous variations in arousal modulate subsequent visual processing and local field potential dynamics in the ferret during quiet wakefulness.

Behavioral states affect neuronal responses throughout the cortex and influence visual processing. Quiet wakefulness (QW) is a behavioral state during which subjects are quiescent but awake and connected to the environment. Here, we examined the effects of pre-stimulus arousal variability on post-stimulus neural activity in the primary visual cortex and posterior parietal cortex in awake ferrets, using pupil diameter as an indicator of arousal. We observed that the power of stimuli-induced alpha (8-12 Hz) decreases when the arousal level increases. The peak of alpha power shifts depending on arousal. High arousal increases inter- and intra-areal coherence. Using a simplified model of laminar circuits, we show that this connectivity pattern is compatible with feedback signals targeting infragranular layers in area posterior parietal cortex and supragranular layers in V1. During high arousal, neurons in V1 displayed higher firing rates at their preferred orientations. Broad-spiking cells in V1 are entrained to high-frequency oscillations (>80 Hz), whereas narrow-spiking neurons are phase-locked to low- (12-18 Hz) and high-frequency (>80 Hz) rhythms. These results indicate that the variability and sensitivity of post-stimulus cortical responses and coherence depend on the pre-stimulus behavioral state and account for the neuronal response variability observed during repeated stimulation.

Mechanisms of distributed working memory in a large-scale network of macaque neocortex.

Neural activity underlying working memory is not a local phenomenon but distributed across multiple brain regions. To elucidate the circuit mechanism of such distributed activity, we developed an anatomically constrained computational model of large-scale macaque cortex. We found that mnemonic internal states may emerge from inter-areal reverberation, even in a regime where none of the isolated areas is capable of generating self-sustained activity. The mnemonic activity pattern along the cortical hierarchy indicates a transition in space, separating areas engaged in working memory and those which do not. A host of spatially distinct attractor states is found, potentially subserving various internal processes. The model yields testable predictions, including the idea of counterstream inhibitory bias, the role of prefrontal areas in controlling distributed attractors, and the resilience of distributed activity to lesions or inactivation. This work provides a theoretical framework for identifying large-scale brain mechanisms and computational principles of distributed cognitive processes.

The continued need for animals to advance brain research.

Policymakers aim to move toward animal-free alternatives for scientific research and have introduced very strict regulations for animal research. We argue that, for neuroscience research, until viable and translational alternatives become available and the value of these alternatives has been proven, the use of animals should not be compromised.

Cerebellar Purkinje cells can differentially modulate coherence between sensory and motor cortex depending on region and behavior.

Activity of sensory and motor cortices is essential for sensorimotor integration. In particular, coherence between these areas may indicate binding of critical functions like perception, motor planning, action, or sleep. Evidence is accumulating that cerebellar output modulates cortical activity and coherence, but how, when, and where it does so is unclear. We studied activity in and coherence between S1 and M1 cortices during whisker stimulation in the absence and presence of optogenetic Purkinje cell stimulation in crus 1 and 2 of awake mice, eliciting strong simple spike rate modulation. Without Purkinje cell stimulation, whisker stimulation triggers fast responses in S1 and M1 involving transient coherence in a broad spectrum. Simultaneous stimulation of Purkinje cells and whiskers affects amplitude and kinetics of sensory responses in S1 and M1 and alters the estimated S1-M1 coherence in theta and gamma bands, allowing bidirectional control dependent on behavioral context. These effects are absent when Purkinje cell activation is delayed by 20 ms. Focal stimulation of Purkinje cells revealed site specificity, with cells in medial crus 2 showing the most prominent and selective impact on estimated coherence, i.e., a strong suppression in the gamma but not the theta band. Granger causality analyses and computational modeling of the involved networks suggest that Purkinje cells control S1-M1 phase consistency predominantly via ventrolateral thalamus and M1. Our results indicate that activity of sensorimotor cortices can be dynamically and functionally modulated by specific cerebellar inputs, highlighting a widespread role of the cerebellum in coordinating sensorimotor behavior.

Directed functional and structural connectivity in a large-scale model for the mouse cortex.

Inferring the structural connectivity from electrophysiological measurements is a fundamental challenge in systems neuroscience. Directed functional connectivity measures, such as the generalized partial directed coherence (GPDC), provide estimates of the causal influence between areas. However, the relation between causality estimates and structural connectivity is still not clear. We analyzed this problem by evaluating the effectiveness of GPDC to estimate the connectivity of a ground-truth, data-constrained computational model of a large-scale network model of the mouse cortex. The model contains 19 cortical areas composed of spiking neurons, with areas connected by long-range projections with weights obtained from a tract-tracing cortical connectome. We show that GPDC values provide a reasonable estimate of structural connectivity, with an average Pearson correlation over simulations of 0.74. Moreover, even in a typical electrophysiological recording scenario containing five areas, the mean correlation was above 0.6. These results suggest that it may be possible to empirically estimate structural connectivity from functional connectivity even when detailed whole-brain recordings are not achievable.

Neural Correlates of Multisensory Detection Behavior: Comparison of Primary and Higher-Order Visual Cortex.

We act upon stimuli in our surrounding environment by gathering the multisensory information they convey and by integrating this information to decide on a behavioral action. We hypothesized that the anterolateral secondary visual cortex (area AL) of the mouse brain may serve as a hub for sensorimotor transformation of audiovisual information. We imaged neuronal activity in primary visual cortex (V1) and AL of the mouse during a detection task using visual, auditory, and audiovisual stimuli. We found that AL neurons were more sensitive to weak uni- and multisensory stimuli compared to V1. Depending on contrast, different subsets of AL and V1 neurons showed cross-modal modulation of visual responses. During audiovisual stimulation, AL neurons showed stronger differentiation of behaviorally reported versus unreported stimuli compared to V1, whereas V1 showed this distinction during unisensory visual stimulation. Thus, neural population activity in area AL correlates more closely with multisensory detection behavior than V1.

Engagement of Pulvino-cortical Feedforward and Feedback Pathways in Cognitive Computations.

Computational modeling of brain mechanisms of cognition has largely focused on the cortex, but recent experiments have shown that higher-order nuclei of the thalamus participate in major cognitive functions and are implicated in psychiatric disorders. Here, we show that a pulvino-cortical circuit model, composed of the pulvinar and two cortical areas, captures several physiological and behavioral observations related to the macaque pulvinar. Effective connections between the two cortical areas are gated by the pulvinar, allowing the pulvinar to shift the operation regime of these areas during attentional processing and working memory and resolve conflict in decision making. Furthermore, cortico-pulvinar projections that engage the thalamic reticular nucleus enable the pulvinar to estimate decision confidence. Finally, feedforward and feedback pulvino-cortical pathways participate in frequency-dependent inter-areal interactions that modify the relative hierarchical positions of cortical areas. Overall, our model suggests that the pulvinar provides crucial contextual modulation to cortical computations associated with cognition.

Inter-areal Balanced Amplification Enhances Signal Propagation in a Large-Scale Circuit Model of the Primate Cortex.

Understanding reliable signal transmission represents a notable challenge for cortical systems, which display a wide range of weights of feedforward and feedback connections among heterogeneous areas. We re-examine the question of signal transmission across the cortex in a network model based on mesoscopic directed and weighted inter-areal connectivity data of the macaque cortex. Our findings reveal that, in contrast to purely feedforward propagation models, the presence of long-range excitatory feedback projections could compromise stable signal propagation. Using population rate models as well as a spiking network model, we find that effective signal propagation can be accomplished by balanced amplification across cortical areas while ensuring dynamical stability. Moreover, the activation of prefrontal cortex in our model requires the input strength to exceed a threshold, which is consistent with the ignition model of conscious processing. These findings demonstrate our model as an anatomically realistic platform for investigations of global primate cortex dynamics.

Paradoxical response reversal of top-down modulation in cortical circuits with three interneuron types.

Pyramidal cells and interneurons expressing parvalbumin (PV), somatostatin (SST), and vasoactive intestinal peptide (VIP) show cell-type-specific connectivity patterns leading to a canonical microcircuit across cortex. Experiments recording from this circuit often report counterintuitive and seemingly contradictory findings. For example, the response of SST cells in mouse V1 to top-down behavioral modulation can change its sign when the visual input changes, a phenomenon that we call response reversal. We developed a theoretical framework to explain these seemingly contradictory effects as emerging phenomena in circuits with two key features: interactions between multiple neural populations and a nonlinear neuronal input-output relationship. Furthermore, we built a cortical circuit model which reproduces counterintuitive dynamics observed in mouse V1. Our analytical calculations pinpoint connection properties critical to response reversal, and predict additional novel types of complex dynamics that could be tested in future experiments.

Nonstationary Stochastic Dynamics Underlie Spontaneous Transitions between Active and Inactive Behavioral States.

The neural basis of spontaneous movement generation is a fascinating open question. Long-term monitoring of fish, swimming freely in a constant sensory environment, has revealed a sequence of behavioral states that alternate randomly and spontaneously between periods of activity and inactivity. We show that key dynamical features of this sequence are captured by a 1-D diffusion process evolving in a nonlinear double well energy landscape, in which a slow variable modulates the relative depth of the wells. This combination of stochasticity, nonlinearity, and nonstationary forcing correctly captures the vastly different timescales of fluctuations observed in the data (∼1 to ∼1000 s), and yields long-tailed residence time distributions (RTDs) also consistent with the data. In fact, our model provides a simple mechanism for the emergence of long-tailed distributions in spontaneous animal behavior. We interpret the stochastic variable of this dynamical model as a decision-like variable that, upon reaching a threshold, triggers the transition between states. Our main finding is thus the identification of a threshold crossing process as the mechanism governing spontaneous movement initiation and termination, and to infer the presence of underlying nonstationary agents. Another important outcome of our work is a dimensionality reduction scheme that allows similar segments of data to be grouped together. This is done by first extracting geometrical features in the dataset and then applying principal component analysis over the feature space. Our study is novel in its ability to model nonstationary behavioral data over a wide range of timescales.

Feedforward and feedback frequency-dependent interactions in a large-scale laminar network of the primate cortex.

Interactions between top-down and bottom-up processes in the cerebral cortex hold the key to understanding attentional processes, predictive coding, executive control, and a gamut of other brain functions. However, the underlying circuit mechanism remains poorly understood and represents a major challenge in neuroscience. We approached this problem using a large-scale computational model of the primate cortex constrained by new directed and weighted connectivity data. In our model, the interplay between feedforward and feedback signaling depends on the cortical laminar structure and involves complex dynamics across multiple (intralaminar, interlaminar, interareal, and whole cortex) scales. The model was tested by reproducing, as well as providing insights into, a wide range of neurophysiological findings about frequency-dependent interactions between visual cortical areas, including the observation that feedforward pathways are associated with enhanced gamma (30 to 70 Hz) oscillations, whereas feedback projections selectively modulate alpha/low-beta (8 to 15 Hz) oscillations. Furthermore, the model reproduces a functional hierarchy based on frequency-dependent Granger causality analysis of interareal signaling, as reported in recent monkey and human experiments, and suggests a mechanism for the observed context-dependent hierarchy dynamics. Together, this work highlights the necessity of multiscale approaches and provides a modeling platform for studies of large-scale brain circuit dynamics and functions.

Differential effects of excitatory and inhibitory heterogeneity on the gain and asynchronous state of sparse cortical networks.

Recent experimental and theoretical studies have highlighted the importance of cell-to-cell differences in the dynamics and functions of neural networks, such as in different types of neural coding or synchronization. It is still not known, however, how neural heterogeneity can affect cortical computations, or impact the dynamics of typical cortical circuits constituted of sparse excitatory and inhibitory networks. In this work, we analytically and numerically study the dynamics of a typical cortical circuit with a certain level of neural heterogeneity. Our circuit includes realistic features found in real cortical populations, such as network sparseness, excitatory, and inhibitory subpopulations of neurons, and different cell-to-cell heterogeneities for each type of population in the system. We find highly differentiated roles for heterogeneity, depending on the subpopulation in which it is found. In particular, while heterogeneity among excitatory neurons non-linearly increases the mean firing rate and linearizes the f-I curves, heterogeneity among inhibitory neurons may decrease the network activity level and induces divisive gain effects in the f-I curves of the excitatory cells, providing an effective gain control mechanism to influence information flow. In addition, we compute the conditions for stability of the network activity, finding that the synchronization onset is robust to inhibitory heterogeneity, but it shifts to lower input levels for higher excitatory heterogeneity. Finally, we provide an extension of recently reported heterogeneity-induced mechanisms for signal detection under rate coding, and we explore the validity of our findings when multiple sources of heterogeneity are present. These results allow for a detailed characterization of the role of neural heterogeneity in asynchronous cortical networks.

Subtractive, divisive and non-monotonic gain control in feedforward nets linearized by noise and delays.

The control of input-to-output mappings, or gain control, is one of the main strategies used by neural networks for the processing and gating of information. Using a spiking neural network model, we studied the gain control induced by a form of inhibitory feedforward circuitry-also known as "open-loop feedback"-, which has been experimentally observed in a cerebellum-like structure in weakly electric fish. We found, both analytically and numerically, that this network displays three different regimes of gain control: subtractive, divisive, and non-monotonic. Subtractive gain control was obtained when noise is very low in the network. Also, it was possible to change from divisive to non-monotonic gain control by simply modulating the strength of the feedforward inhibition, which may be achieved via long-term synaptic plasticity. The particular case of divisive gain control has been previously observed in vivo in weakly electric fish. These gain control regimes were robust to the presence of temporal delays in the inhibitory feedforward pathway, which were found to linearize the input-to-output mappings (or f-I curves) via a novel variability-increasing mechanism. Our findings highlight the feedforward-induced gain control analyzed here as a highly versatile mechanism of information gating in the brain.

Learning contrast-invariant cancellation of redundant signals in neural systems.

Cancellation of redundant information is a highly desirable feature of sensory systems, since it would potentially lead to a more efficient detection of novel information. However, biologically plausible mechanisms responsible for such selective cancellation, and especially those robust to realistic variations in the intensity of the redundant signals, are mostly unknown. In this work, we study, via in vivo experimental recordings and computational models, the behavior of a cerebellar-like circuit in the weakly electric fish which is known to perform cancellation of redundant stimuli. We experimentally observe contrast invariance in the cancellation of spatially and temporally redundant stimuli in such a system. Our model, which incorporates heterogeneously-delayed feedback, bursting dynamics and burst-induced STDP, is in agreement with our in vivo observations. In addition, the model gives insight on the activity of granule cells and parallel fibers involved in the feedback pathway, and provides a strong prediction on the parallel fiber potentiation time scale. Finally, our model predicts the existence of an optimal learning contrast around 15% contrast levels, which are commonly experienced by interacting fish.

Emergence of resonances in neural systems: the interplay between adaptive threshold and short-term synaptic plasticity.

In this work we study the detection of weak stimuli by spiking (integrate-and-fire) neurons in the presence of certain level of noisy background neural activity. Our study has focused in the realistic assumption that the synapses in the network present activity-dependent processes, such as short-term synaptic depression and facilitation. Employing mean-field techniques as well as numerical simulations, we found that there are two possible noise levels which optimize signal transmission. This new finding is in contrast with the classical theory of stochastic resonance which is able to predict only one optimal level of noise. We found that the complex interplay between adaptive neuron threshold and activity-dependent synaptic mechanisms is responsible for this new phenomenology. Our main results are confirmed by employing a more realistic FitzHugh-Nagumo neuron model, which displays threshold variability, as well as by considering more realistic stochastic synaptic models and realistic signals such as poissonian spike trains.

Irregular dynamics in up and down cortical states.

Complex coherent dynamics is present in a wide variety of neural systems. A typical example is the voltage transitions between up and down states observed in cortical areas in the brain. In this work, we study this phenomenon via a biologically motivated stochastic model of up and down transitions. The model is constituted by a simple bistable rate dynamics, where the synaptic current is modulated by short-term synaptic processes which introduce stochasticity and temporal correlations. A complete analysis of our model, both with mean-field approaches and numerical simulations, shows the appearance of complex transitions between high (up) and low (down) neural activity states, driven by the synaptic noise, with permanence times in the up state distributed according to a power-law. We show that the experimentally observed large fluctuation in up and down permanence times can be explained as the result of sufficiently noisy dynamical synapses with sufficiently large recovery times. Static synapses cannot account for this behavior, nor can dynamical synapses in the absence of noise.