Approximating Nonlinear Functions With Latent Boundaries in Low-Rank Excitatory-Inhibitory Spiking Networks.

Deep feedforward and recurrent neural networks have become successful functional models of the brain, but they neglect obvious biological details such as spikes and Dale's law. Here we argue that these details are crucial in order to understand how real neural circuits operate. Towards this aim, we put forth a new framework for spike-based computation in low-rank excitatory-inhibitory spiking networks. By considering populations with rank-1 connectivity, we cast each neuron's spiking threshold as a boundary in a low-dimensional input-output space. We then show how the combined thresholds of a population of inhibitory neurons form a stable boundary in this space, and those of a population of excitatory neurons form an unstable boundary. Combining the two boundaries results in a rank-2 excitatory-inhibitory (EI) network with inhibition-stabilized dynamics at the intersection of the two boundaries. The computation of the resulting networks can be understood as the difference of two convex functions and is thereby capable of approximating arbitrary non-linear input-output mappings. We demonstrate several properties of these networks, including noise suppression and amplification, irregular activity and synaptic balance, as well as how they relate to rate network dynamics in the limit that the boundary becomes soft. Finally, while our work focuses on small networks (5-50 neurons), we discuss potential avenues for scaling up to much larger networks. Overall, our work proposes a new perspective on spiking networks that may serve as a starting point for a mechanistic understanding of biological spike-based computation.

Dimensionality reduction reveals separate translation and rotation populations in the zebrafish hindbrain.

In many brain areas, neuronal activity is associated with a variety of behavioral and environmental variables. In particular, neuronal responses in the zebrafish hindbrain relate to oculomotor and swimming variables as well as sensory information. However, the precise functional organization of the neurons has been difficult to unravel because neuronal responses are heterogeneous. Here, we used dimensionality reduction methods on neuronal population data to reveal the role of the hindbrain in visually driven oculomotor behavior and swimming. We imaged neuronal activity in zebrafish expressing GCaMP6s in the nucleus of almost all neurons while monitoring the behavioral response to gratings that rotated with different speeds. We then used reduced-rank regression, a method that condenses the sensory and motor variables into a smaller number of "features," to predict the fluorescence traces of all ROIs (regions of interest). Despite the potential complexity of the visuo-motor transformation, our analysis revealed that a large fraction of the population activity can be explained by only two features. Based on the contribution of these features to each ROI's activity, ROIs formed three clusters. One cluster was related to vergent movements and swimming, whereas the other two clusters related to leftward and rightward rotation. Voxels corresponding to these clusters were segregated anatomically, with leftward and rightward rotation clusters located selectively to the left and right hemispheres, respectively. Just as described in many cortical areas, our analysis revealed that single-neuron complexity co-exists with a simpler population-level description, thereby providing insights into the organization of visuo-motor transformations in the hindbrain.

Action suppression reveals opponent parallel control via striatal circuits.

The direct and indirect pathways of the basal ganglia are classically thought to promote and suppress action, respectively1. However, the observed co-activation of striatal direct and indirect medium spiny neurons2 (dMSNs and iMSNs, respectively) has challenged this view. Here we study these circuits in mice performing an interval categorization task that requires a series of self-initiated and cued actions and, critically, a sustained period of dynamic action suppression. Although movement produced the co-activation of iMSNs and dMSNs in the sensorimotor, dorsolateral striatum (DLS), fibre photometry and photo-identified electrophysiological recordings revealed signatures of functional opponency between the two pathways during action suppression. Notably, optogenetic inhibition showed that DLS circuits were largely engaged to suppress-and not promote-action. Specifically, iMSNs on a given hemisphere were dynamically engaged to suppress tempting contralateral action. To understand how such regionally specific circuit function arose, we constructed a computational reinforcement learning model that reproduced key features of behaviour, neural activity and optogenetic inhibition. The model predicted that parallel striatal circuits outside the DLS learned the action-promoting functions, generating the temptation to act. Consistent with this, optogenetic inhibition experiments revealed that dMSNs in the associative, dorsomedial striatum, in contrast to those in the DLS, promote contralateral actions. These data highlight how opponent interactions between multiple circuit- and region-specific basal ganglia processes can lead to behavioural control, and establish a critical role for the sensorimotor indirect pathway in the proactive suppression of tempting actions.

Efficient coding of cognitive variables underlies dopamine response and choice behavior.

Reward expectations based on internal knowledge of the external environment are a core component of adaptive behavior. However, internal knowledge may be inaccurate or incomplete due to errors in sensory measurements. Some features of the environment may also be encoded inaccurately to minimize representational costs associated with their processing. In this study, we investigated how reward expectations are affected by features of internal representations by studying behavior and dopaminergic activity while mice make time-based decisions. We show that several possible representations allow a reinforcement learning agent to model animals' overall performance during the task. However, only a small subset of highly compressed representations simultaneously reproduced the co-variability in animals' choice behavior and dopaminergic activity. Strikingly, these representations predict an unusual distribution of response times that closely match animals' behavior. These results inform how constraints of representational efficiency may be expressed in encoding representations of dynamic cognitive variables used for reward-based computations.

The geometry of robustness in spiking neural networks.

Neural systems are remarkably robust against various perturbations, a phenomenon that still requires a clear explanation. Here, we graphically illustrate how neural networks can become robust. We study spiking networks that generate low-dimensional representations, and we show that the neurons' subthreshold voltages are confined to a convex region in a lower-dimensional voltage subspace, which we call a 'bounding box'. Any changes in network parameters (such as number of neurons, dimensionality of inputs, firing thresholds, synaptic weights, or transmission delays) can all be understood as deformations of this bounding box. Using these insights, we show that functionality is preserved as long as perturbations do not destroy the integrity of the bounding box. We suggest that the principles underlying robustness in these networks - low-dimensional representations, heterogeneity of tuning, and precise negative feedback - may be key to understanding the robustness of neural systems at the circuit level.

Feedforward and feedback interactions between visual cortical areas use different population activity patterns.

Brain function relies on the coordination of activity across multiple, recurrently connected brain areas. For instance, sensory information encoded in early sensory areas is relayed to, and further processed by, higher cortical areas and then fed back. However, the way in which feedforward and feedback signaling interact with one another is incompletely understood. Here we investigate this question by leveraging simultaneous neuronal population recordings in early and midlevel visual areas (V1-V2 and V1-V4). Using a dimensionality reduction approach, we find that population interactions are feedforward-dominated shortly after stimulus onset and feedback-dominated during spontaneous activity. The population activity patterns most correlated across areas were distinct during feedforward- and feedback-dominated periods. These results suggest that feedforward and feedback signaling rely on separate "channels", which allows feedback signals to not directly affect activity that is fed forward.

Disentangling the flow of signals between populations of neurons.

Technological advances now allow us to record from large populations of neurons across multiple brain areas. These recordings may illuminate how communication between areas contributes to brain function, yet a substantial barrier remains: how do we disentangle the concurrent, bidirectional flow of signals between populations of neurons? We propose here a dimensionality reduction framework, delayed latents across groups (DLAG), that disentangles signals relayed in each direction, identifies how these signals are represented by each population and characterizes how they evolve within and across trials. We demonstrate that DLAG performs well on synthetic datasets similar in scale to current neurophysiological recordings. Then we study simultaneously recorded populations in primate visual areas V1 and V2, where DLAG reveals signatures of bidirectional yet selective communication. Our framework lays a foundation for dissecting the intricate flow of signals across populations of neurons, and how this signalling contributes to cortical computation.

Representational geometry of perceptual decisions in the monkey parietal cortex.

Lateral intraparietal (LIP) neurons represent formation of perceptual decisions involving eye movements. In circuit models for these decisions, neural ensembles that encode actions compete to form decisions. Consequently, representation and readout of the decision variables (DVs) are implemented similarly for decisions with identical competing actions, irrespective of input and task context differences. Further, DVs are encoded as partially potentiated action plans through balance of activity of action-selective ensembles. Here, we test those core principles. We show that in a novel face-discrimination task, LIP firing rates decrease with supporting evidence, contrary to conventional motion-discrimination tasks. These opposite response patterns arise from similar mechanisms in which decisions form along curved population-response manifolds misaligned with action representations. These manifolds rotate in state space based on context, indicating distinct optimal readouts for different tasks. We show similar manifolds in lateral and medial prefrontal cortices, suggesting similar representational geometry across decision-making circuits.

Statistical methods for dissecting interactions between brain areas.

The brain is composed of many functionally distinct areas. This organization supports distributed processing, and requires the coordination of signals across areas. Our understanding of how populations of neurons in different areas interact with each other is still in its infancy. As the availability of recordings from large populations of neurons across multiple brain areas increases, so does the need for statistical methods that are well suited for dissecting and interrogating these recordings. Here we review multivariate statistical methods that have been, or could be, applied to this class of recordings. By leveraging population responses, these methods can provide a rich description of inter-areal interactions. At the same time, these methods can introduce interpretational challenges. We thus conclude by discussing how to interpret the outputs of these methods to further our understanding of inter-areal interactions.

Principles of Corticocortical Communication: Proposed Schemes and Design Considerations.

Nearly all brain functions involve routing neural activity among a distributed network of areas. Understanding this routing requires more than a description of interareal anatomical connectivity: it requires understanding what controls the flow of signals through interareal circuitry and how this communication might be modulated to allow flexible behavior. Here we review proposals of how communication, particularly between visual cortical areas, is instantiated and modulated, highlighting recent work that offers new perspectives. We suggest transitioning from a focus on assessing changes in the strength of interareal interactions, as often seen in studies of interareal communication, to a broader consideration of how different signaling schemes might contribute to computation. To this end, we discuss a set of features that might be desirable for a communication scheme.

Learning to represent signals spike by spike.

Networks based on coordinated spike coding can encode information with high efficiency in the spike trains of individual neurons. These networks exhibit single-neuron variability and tuning curves as typically observed in cortex, but paradoxically coincide with a precise, non-redundant spike-based population code. However, it has remained unclear whether the specific synaptic connectivities required in these networks can be learnt with local learning rules. Here, we show how to learn the required architecture. Using coding efficiency as an objective, we derive spike-timing-dependent learning rules for a recurrent neural network, and we provide exact solutions for the networks' convergence to an optimal state. As a result, we deduce an entire network from its input distribution and a firing cost. After learning, basic biophysical quantities such as voltages, firing thresholds, excitation, inhibition, or spikes acquire precise functional interpretations.

Decoding and encoding (de)mixed population responses.

A central tenet of neuroscience is that the brain works through large populations of interacting neurons. With recent advances in recording techniques, the inner working of these populations has come into full view. Analyzing the resulting large-scale data sets is challenging because of the often complex and 'mixed' dependency of neural activities on experimental parameters, such as stimuli, decisions, or motor responses. Here we review recent insights gained from analyzing these data with dimensionality reduction methods that 'demix' these dependencies. We demonstrate that the mappings from (carefully chosen) experimental parameters to population activities appear to be typical and stable across tasks, brain areas, and animals, and are often identifiable by linear methods. By considering when and why dimensionality reduction and demixing work well, we argue for a view of population coding in which populations represent (demixed) latent signals, corresponding to stimuli, decisions, motor responses, and so on. These latent signals are encoded into neural population activity via non-linear mappings and decoded via linear readouts. We explain how such a scheme can facilitate the propagation of information across cortical areas, and we review neural network architectures that can reproduce the encoding and decoding of latent signals in population activities. These architectures promise a link from the biophysics of single neurons to the activities of neural populations.

State-dependent geometry of population activity in rat auditory cortex.

The accuracy of the neural code depends on the relative embedding of signal and noise in the activity of neural populations. Despite a wealth of theoretical work on population codes, there are few empirical characterizations of the high-dimensional signal and noise subspaces. We studied the geometry of population codes in the rat auditory cortex across brain states along the activation-inactivation continuum, using sounds varying in difference and mean level across the ears. As the cortex becomes more activated, single-hemisphere populations go from preferring contralateral loud sounds to a symmetric preference across lateralizations and intensities, gain-modulation effectively disappears, and the signal and noise subspaces become approximately orthogonal to each other and to the direction corresponding to global activity modulations. Level-invariant decoding of sound lateralization also becomes possible in the active state. Our results provide an empirical foundation for the geometry and state-dependence of cortical population codes.

Cortical Areas Interact through a Communication Subspace.

Most brain functions involve interactions among multiple, distinct areas or nuclei. For instance, visual processing in primates requires the appropriate relaying of signals across many distinct cortical areas. Yet our understanding of how populations of neurons in interconnected brain areas communicate is in its infancy. Here we investigate how trial-to-trial fluctuations of population responses in primary visual cortex (V1) are related to simultaneously recorded population responses in area V2. Using dimensionality reduction methods, we find that V1-V2 interactions occur through a communication subspace: V2 fluctuations are related to a small subset of V1 population activity patterns, distinct from the largest fluctuations shared among neurons within V1. In contrast, interactions between subpopulations within V1 are less selective. We propose that the communication subspace may be a general, population-level mechanism by which activity can be selectively routed across brain areas.

Optimal compensation for neuron loss.

The brain has an impressive ability to withstand neural damage. Diseases that kill neurons can go unnoticed for years, and incomplete brain lesions or silencing of neurons often fail to produce any behavioral effect. How does the brain compensate for such damage, and what are the limits of this compensation? We propose that neural circuits instantly compensate for neuron loss, thereby preserving their function as much as possible. We show that this compensation can explain changes in tuning curves induced by neuron silencing across a variety of systems, including the primary visual cortex. We find that compensatory mechanisms can be implemented through the dynamics of networks with a tight balance of excitation and inhibition, without requiring synaptic plasticity. The limits of this compensatory mechanism are reached when excitation and inhibition become unbalanced, thereby demarcating a recovery boundary, where signal representation fails and where diseases may become symptomatic.

Distributed and Mixed Information in Monosynaptic Inputs to Dopamine Neurons.

Dopamine neurons encode the difference between actual and predicted reward, or reward prediction error (RPE). Although many models have been proposed to account for this computation, it has been difficult to test these models experimentally. Here we established an awake electrophysiological recording system, combined with rabies virus and optogenetic cell-type identification, to characterize the firing patterns of monosynaptic inputs to dopamine neurons while mice performed classical conditioning tasks. We found that each variable required to compute RPE, including actual and predicted reward, was distributed in input neurons in multiple brain areas. Further, many input neurons across brain areas signaled combinations of these variables. These results demonstrate that even simple arithmetic computations such as RPE are not localized in specific brain areas but, rather, distributed across multiple nodes in a brain-wide network. Our systematic method to examine both activity and connectivity revealed unexpected redundancy for a simple computation in the brain.

Demixed principal component analysis of neural population data.

Neurons in higher cortical areas, such as the prefrontal cortex, are often tuned to a variety of sensory and motor variables, and are therefore said to display mixed selectivity. This complexity of single neuron responses can obscure what information these areas represent and how it is represented. Here we demonstrate the advantages of a new dimensionality reduction technique, demixed principal component analysis (dPCA), that decomposes population activity into a few components. In addition to systematically capturing the majority of the variance of the data, dPCA also exposes the dependence of the neural representation on task parameters such as stimuli, decisions, or rewards. To illustrate our method we reanalyze population data from four datasets comprising different species, different cortical areas and different experimental tasks. In each case, dPCA provides a concise way of visualizing the data that summarizes the task-dependent features of the population response in a single figure.

Efficient codes and balanced networks.

Recent years have seen a growing interest in inhibitory interneurons and their circuits. A striking property of cortical inhibition is how tightly it balances excitation. Inhibitory currents not only match excitatory currents on average, but track them on a millisecond time scale, whether they are caused by external stimuli or spontaneous fluctuations. We review, together with experimental evidence, recent theoretical approaches that investigate the advantages of such tight balance for coding and computation. These studies suggest a possible revision of the dominant view that neurons represent information with firing rates corrupted by Poisson noise. Instead, tight excitatory/inhibitory balance may be a signature of a highly cooperative code, orders of magnitude more precise than a Poisson rate code. Moreover, tight balance may provide a template that allows cortical neurons to construct high-dimensional population codes and learn complex functions of their inputs.

On the number of neurons and time scale of integration underlying the formation of percepts in the brain.

All of our perceptual experiences arise from the activity of neural populations. Here we study the formation of such percepts under the assumption that they emerge from a linear readout, i.e., a weighted sum of the neurons' firing rates. We show that this assumption constrains the trial-to-trial covariance structure of neural activities and animal behavior. The predicted covariance structure depends on the readout parameters, and in particular on the temporal integration window w and typical number of neurons K used in the formation of the percept. Using these predictions, we show how to infer the readout parameters from joint measurements of a subject's behavior and neural activities. We consider three such scenarios: (1) recordings from the complete neural population, (2) recordings of neuronal sub-ensembles whose size exceeds K, and (3) recordings of neuronal sub-ensembles that are smaller than K. Using theoretical arguments and artificially generated data, we show that the first two scenarios allow us to recover the typical spatial and temporal scales of the readout. In the third scenario, we show that the readout parameters can only be recovered by making additional assumptions about the structure of the full population activity. Our work provides the first thorough interpretation of (feed-forward) percept formation from a population of sensory neurons. We discuss applications to experimental recordings in classic sensory decision-making tasks, which will hopefully provide new insights into the nature of perceptual integration.