In 1949, Donald Hebb proposed that neuronal assemblies with temporally specific patterns of activity form the building blocks of perception, cognition, and behavior. Finding the structural underpinning of such assemblies has been technically challenging due to a lack of large-scale structure-activity maps. Here, we combine in vivo optical physiology with postmortem electron microscopy (EM) in the same tissue volume. Using higher-order correlations in fluorescence traces, we extract neuronal assemblies. Physiologically, we show that these assemblies respond more reliably to repeated natural movies than size-matched control ensembles and decode such stimuli more accurately. Structurally, we find that over a quarter of the pyramidal neurons do not participate in any assembly and are significantly less integrated into the connectome than those that do. We do not observe a marked increase in the strength of monosynaptic excitatory connections between neurons sharing assembly assignment, but instead find significantly stronger indirect inhibitory connections targeting cells in other assemblies. These results show that assemblies can serve as functional units of perception and suggest they may be structurally delineated by mutual inhibition.

The simple linear threshold units used in many artificial neural networks have a limited computational capacity. Famously, a single unit cannot handle non-linearly separable problems like XOR. In contrast, real neurons exhibit complex morphologies as well as active dendritic integration, suggesting that their computational capacities outperform those of simple linear units. Considering specific families of Boolean functions, we empirically examine the computational limits of single units that incorporate more complex dendritic structures. For random Boolean functions, we show that there is a phase transition in learnability as a function of the input dimension, with most random functions below a certain critical dimension being learnable and those above not. This critical dimension is best predicted by the overall size of the dendritic arbor. This demonstrates that real neurons have a far higher computational complexity than is usually considered in neural models, whether in machine learning or computational neuroscience. Furthermore, using architectures that are, respectively, more "apical" or "basal" we show that there are non-trivially disjoint sets of learnable functions by each type of neuron. Importantly, these two types of architectures differ in the robustness and generality of the computations they can perform. The basal-like architecture shows a higher probability of function realization, while the apical-like architecture shows an advantage with fast retraining for different functions. Given the cell-type specificity of morphological characteristics, these results suggest both that different components of the dendritic arbor as well as distinct cell types may have distinct computational roles. In single neurons, morphology sculpts computation, shaping not only what neurons do, but how they learn and adapt. Our analysis offers new directions for neuron-level inductive biases in NeuroAI models using scalable models for neuronal cell-type specific computation.

Influenced by non-stationary factors such as brain states and behavior, neurons exhibit substantial response variability even to identical stimuli. However, it remains unclear how their relative impact on neuronal variability evolves over time. To address this question, we designed an encoding model conditioned on latent states to partition variability in the mouse visual cortex across internal brain dynamics, behavior, and external visual stimulus. Applying a hidden Markov model to local field potentials, we consistently identified three distinct oscillation states, each with a unique variability profile. Regression models within each state revealed a dynamic composition of factors influencing spiking variability, with the dominant factor switching within seconds. The state-conditioned regression model uncovered extensive diversity in source contributions across units, varying in accordance with anatomical hierarchy and internal state. This heterogeneity in encoding underscores the importance of partitioning variability over time, particularly when considering the influence of non-stationary factors on sensory processing.

Self-assessment of one’s choices, i.e., confidence, is the topic of many decision neuroscience studies. Computational models of confidence, however, are limited to specific scenarios such as between choices with the same value. Here we present a normative framework for modeling decision confidence that is generalizable to various tasks and experimental setups. We further drive the implications of our model from both theoretical and experimental points of view. Specifically, we show that our model maps to the planning as an inference framework where the objective function is maximizing the gained reward and information entropy of the policy. Moreover, we validate our model on two different psychophysics experiments and show its superiority over other approaches in explaining subjects' confidence reports.

The diverse functions of different cortical areas are thought to arise from their distinct groups of inputs. However, additional organizing principles may exist in the spatial structure of converging inputs. We investigated spatial convergence patterns of projections from primary sensory areas to other areas throughout the mouse isocortex. We used a large tract tracing dataset to estimate the dimension of the space into which topographical connections from multiple modalities converged within each other cortical area. We call this measure the topography dimension (TD). TD is higher for areas that receive inputs of similar strength from multiple sensory modalities, and lower when multiple inputs terminate in register with one another. Across the isocortex, TD varied by a factor of ~4. TD was positively correlated with hierarchy score, an independent measure that is based on laminar connection patterns. Furthermore, TD (an anatomical measure) was significantly related to several measures of neural activity. In particular, higher TD was associated with higher neural activity dimension, lower population sparseness, and lower lifetime sparseness of spontaneous activity, independent of an area's hierarchical position. Finally, we analyzed factors that limited TD and found that linear correlations among projections from different areas typically had little impact, while diversity of connection strengths, both between different projections onto the same area, and within projections across different parts of an area, limited TD substantially. This analysis revealed additional intricacy of cortical networks, beyond areas' sets of connections and hierarchical organization. We propose a means of approximating this organization in deep-network models. ### Competing Interest Statement The authors have declared no competing interest.

Deep convolutional neural networks are important models of the visual cortex that account relatively well for brain activity and are able to perform ethologically relevant functions. However, it is unknown which combination of factors, such as network architecture, training objectives, and data best align this family of models with the brain. Here we investigate the statistics of training data. We hypothesized that stimuli that are naturalistic for mice would lead to higher similarity between deep network models and activity in mouse visual cortex. We used a video-game engine to create training datasets in which we varied the naturalism of the environment, the movement statistics, and the optics of the modelled eye. The naturalistic environment substantially and consistently led to greater brain similarity, while the other factors had more subtle and area-specific effects. We then hypothesized that differences in brain similarity between the two environments arose due to differences in spatial frequency spectra, distributions of color and orientation, and/or temporal autocorrelations. To test this, we created abstract environments, composed of cubes and spheres, that resembled the naturalistic and non-naturalistic environments in these respects. Contrary to our expectations, these factors accounted poorly for differences in brain similarity due to the naturalistic and non-naturalistic environments. This suggests that the higher brain similarities we observed after training with the naturalistic environment were due to more complex factors. ### Competing Interest Statement The authors have declared no competing interest.

We present a computationally efficient framework to model a wide range of population structures with high order correlations and a large number of neurons. Our method is based on a special type of Bayesian network that has linear inference time and is founded upon the concept of contextual independence. Our framework is both fast and accurate in approximating neural population structures. Furthermore, our approach enables us to reliably quantify higher order neural correlations. We test our method on simulated neural populations commonly used to generate higher order correlations, as well as on publicly available large-scale neural recordings from the Allen Brain Observatory. Our approach significantly outperforms other models both in terms of statistical measures and alignment with experimental evidence.

A major goal of computational neuroscience is to build accurate models of the activity of neurons that can be used to interpret their function in circuits. Here, we explore using functional cell types to refine single-cell models by grouping them into functionally relevant classes. Formally, we define a hierarchical generative model for cell types, single-cell parameters, and neural responses, and then derive an expectation-maximization algorithm with variational inference that maximizes the likelihood of the neural recordings. We apply this "simultaneous" method to estimate cell types and fit single-cell models from simulated data, and find that it accurately recovers the ground truth parameters. We then apply our approach to in vitro neural recordings from neurons in mouse primary visual cortex, and find that it yields improved prediction of single-cell activity. We demonstrate that the discovered cell-type clusters are well separated and generalizable, and thus amenable to interpretation. We then compare discovered cluster memberships with locational, morphological, and transcriptomic data. Our findings reveal the potential to improve models of neural responses by explicitly allowing for shared functional properties across neurons.

The detection of novel stimuli is critical to learn and survive in a dynamic environment. Though novel stimuli powerfully affect brain activity, their impact on specific cell types and circuits is not well understood. Disinhibition is one candidate mechanism for novelty-induced enhancements in activity. Here we characterize the impact of stimulus novelty on disinhibitory circuit components using longitudinal 2-photon calcium imaging of Vip, Sst, and excitatory populations in the mouse visual cortex. Mice learn a behavioral task with stimuli that become highly familiar, then are tested on both familiar and novel stimuli. Mice consistently perform the task with novel stimuli, yet responses to stimulus presentations and stimulus omissions are dramatically altered. Further, we find that novelty modifies coding of visual as well as behavioral and task information. At the population level, the direction of these changes is consistent with engagement of the Vip-Sst disinhibitory circuit. At the single cell level, we identify separate clusters of Vip, Sst, and excitatory cells with unique patterns of novelty-induced coding changes. This study and the accompanying open-access dataset reveals the impact of novelty on sensory and behavioral representations in visual cortical circuits and establishes novelty as a key driver of cellular functional diversity.

Rapid saccadic eye movements are used by animals to sample different parts of the visual scene. Previous work has investigated neural correlates of these saccades in visual cortical areas such as V1; however, how saccade-responsive neurons are distributed across visual areas, cell types, and cortical layers has remained unknown. Through analyzing 818 1 h experimental sessions from the Allen Brain Observatory, we present a large-scale analysis of saccadic behaviors in head-fixed mice and their neural correlates. We find that saccade-responsive neurons are present across visual cortex, but their distribution varies considerably by transgenically defined cell type, cortical area, and cortical layer. We also find that saccade-responsive neurons do not exhibit distinct visual response properties from the broader neural population, suggesting that the saccadic responses of these neurons are likely not predominantly visually driven. These results provide insight into the roles played by different cell types within a broader, distributed network of sensory and motor interactions.

Not only have deep networks become standard in machine learning, they are increasingly of interest in neuroscience as models of cortical computation that capture relationships between structural and functional properties. In addition they are a useful target of theoretical research into the properties of network computation. Deep networks typically have a serial or approximately serial organization across layers, and this is often mirrored in models that purport to represent computation in mammalian brains. There are, however, multiple examples of parallel pathways in mammalian brains. In some cases, such as the mouse, the entire visual system appears arranged in a largely parallel, rather than serial fashion. While these pathways may be formed by differing cost functions that drive different computations, here we present a new mathematical analysis of learning dynamics in networks that have parallel computational pathways driven by the same cost function. We use the approximation of deep linear networks with large hidden layer sizes to show that, as the depth of the parallel pathways increases, different features of the training set (defined by the singular values of the input-output correlation) will typically concentrate in one of the pathways. This result is derived analytically and demonstrated with numerical simulation with both linear and non-linear networks. Thus, rather than sharing stimulus and task features across multiple pathways, parallel network architectures learn to produce sharply diversified representations with specialized and specific pathways, a mechanism which may hold important consequences for codes in both biological and artificial systems.

Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in mammals, 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.

The anatomy of the mammalian visual system, from the retina to the neocortex, is organized hierarchically1. However, direct observation of cellular-level functional interactions across this hierarchy is lacking due to the challenge of simultaneously recording activity across numerous regions. Here we describe a large, open dataset-part of the Allen Brain Observatory2-that surveys spiking from tens of thousands of units in six cortical and two thalamic regions in the brains of mice responding to a battery of visual stimuli. Using cross-correlation analysis, we reveal that the organization of inter-area functional connectivity during visual stimulation mirrors the anatomical hierarchy from the Allen Mouse Brain Connectivity Atlas3. We find that four classical hierarchical measures-response latency, receptive-field size, phase-locking to drifting gratings and response decay timescale-are all correlated with the hierarchy. Moreover, recordings obtained during a visual task reveal that the correlation between neural activity and behavioural choice also increases along the hierarchy. Our study provides a foundation for understanding coding and signal propagation across hierarchically organized cortical and thalamic visual areas.

How well do deep neural networks fare as models of mouse visual cortex? A majority of research to date suggests results far more mixed than those produced in the modeling of primate visual cortex. Here, we perform a large-scale benchmarking of dozens of deep neural network models in mouse visual cortex with both representational similarity analysis and neural regression. Using the Allen Brain Observatory's 2-photon calcium-imaging dataset of activity in over 6,000 reliable rodent visual cortical neurons recorded in response to natural scenes, we replicate previous findings and resolve previous discrepancies, ultimately demonstrating that modern neural networks can in fact be used to explain activity in the mouse visual cortex to a more reasonable degree than previously suggested. Using our benchmark as an atlas, we offer preliminary answers to overarching questions about levels of analysis (e.g. do models that better predict the representations of individual neurons also predict representational similarity across neural populations?); questions about the properties of models that best predict the visual system overall (e.g. is convolution or category-supervision necessary to better predict neural activity?); and questions about the mapping between biological and artificial representations (e.g. does the information processing hierarchy in deep nets match the anatomical hierarchy of mouse visual cortex?). Along the way, we catalogue a number of models (including vision transformers, MLP-Mixers, normalization free networks, Taskonomy encoders and self-supervised models) outside the traditional circuit of convolutional object recognition. Taken together, our results provide a reference point for future ventures in the deep neural network modeling of mouse visual cortex, hinting at novel combinations of mapping method, architecture, and task to more fully characterize the computational motifs of visual representation in a species so central to neuroscience, but with a perceptual physiology and ecology markedly different from the ones we study in primates.

Extracellular electrophysiology and two-photon calcium imaging are widely used methods for measuring physiological activity with single-cell resolution across large populations of cortical neurons. While each of these two modalities has distinct advantages and disadvantages, neither provides complete, unbiased information about the underlying neural population. Here, we compare evoked responses in visual cortex recorded in awake mice under highly standardized conditions using either imaging of genetically expressed GCaMP6f or electrophysiology with silicon probes. Across all stimulus conditions tested, we observe a larger fraction of responsive neurons in electrophysiology and higher stimulus selectivity in calcium imaging, which was partially reconciled by applying a spikes-to-calcium forward model to the electrophysiology data. However, the forward model could only reconcile differences in responsiveness when restricted to neurons with low contamination and an event rate above a minimum threshold. This work established how the biases of these two modalities impact functional metrics that are fundamental for characterizing sensory-evoked responses.

Fluorescent calcium indicators are often used to investigate neural dynamics, but the relationship between fluorescence and action potentials (APs) remains unclear. Most APs can be detected when the soma almost fills the microscope’s field of view, but calcium indicators are used to image populations of neurons, necessitating a large field of view, generating fewer photons per neuron, and compromising AP detection. Here, we characterized the AP-fluorescence transfer function in vivo for 48 layer 2/3 pyramidal neurons in primary visual cortex, with simultaneous calcium imaging and cell-attached recordings from transgenic mice expressing GCaMP6s or GCaMP6f. While most APs were detected under optimal conditions, under conditions typical of population imaging studies, only a minority of 1 AP and 2 AP events were detected (often <10% and ~20–30%, respectively), emphasizing the limits of AP detection under more realistic imaging conditions.

Biological synaptic plasticity exhibits nonlinearities that are not accounted for by classic Hebbian learning rules. Here, we introduce a simple family of generalized nonlinear Hebbian learning rules. We study the computations implemented by their dynamics in the simple setting of a neuron receiving feedforward inputs. These nonlinear Hebbian rules allow a neuron to learn tensor decompositions of its higher-order input correlations. The particular input correlation decomposed and the form of the decomposition depend on the location of nonlinearities in the plasticity rule. For simple, biologically motivated parameters, the neuron learns eigenvectors of higher-order input correlation tensors. We prove that tensor eigenvectors are attractors and determine their basins of attraction. We calculate the volume of those basins, showing that the dominant eigenvector has the largest basin of attraction. We then study arbitrary learning rules and find that any learning rule that admits a finite Taylor expansion into the neural input and output also has stable equilibria at generalized eigenvectors of higher-order input correlation tensors. Nonlinearities in synaptic plasticity thus allow a neuron to encode higher-order input correlations in a simple fashion.

Neural computation is determined by neurons' dynamics and circuit connectivity. Uncertain and dynamic environments may require neural hardware to adapt to different computational tasks, each requiring different connectivity configurations. At the same time, connectivity is subject to a variety of constraints, placing limits on the possible computations a given neural circuit can perform. Here we examine the hypothesis that the organization of neural circuitry favors computational flexibility: that it makes many computational solutions available, given physiological constraints. From this hypothesis, we develop models of connectivity degree distributions based on constraints on a neuron's total synaptic weight. To test these models, we examine reconstructions of the mushroom bodies from the first instar larva and adultDrosophila melanogaster. We perform a Bayesian model comparison for two constraint models and a random wiring null model. Overall, we find that flexibility under a homeostatically fixed total synaptic weight describes Kenyon cell connectivity better than other models, suggesting a principle shaping the apparently random structure of Kenyon cell wiring. Furthermore, we find evidence that larval Kenyon cells are more flexible earlier in development, suggesting a mechanism whereby neural circuits begin as flexible systems that develop into specialized computational circuits.

Vasoactive intestinal peptide-expressing (VIP) interneurons in the cortex regulate feedback inhibition of pyramidal neurons through suppression of somatostatin-expressing (SST) interneurons and, reciprocally, SST neurons inhibit VIP neurons. Although VIP neuron activity in the primary visual cortex (V1) of mouse is highly correlated with locomotion, the relevance of locomotion-related VIP neuron activity to visual coding is not known. Here we show that VIP neurons in mouse V1 respond strongly to low contrast front-to-back motion that is congruent with self-motion during locomotion but are suppressed by other directions and contrasts. VIP and SST neurons have complementary contrast tuning. Layer 2/3 contains a substantially larger population of low contrast preferring pyramidal neurons than deeper layers, and layer 2/3 (but not deeper layer) pyramidal neurons show bias for front-to-back motion specifically at low contrast. Network modeling indicates that VIP-SST mutual antagonism regulates the gain of the cortex to achieve sensitivity to specific weak stimuli without compromising network stability.

Convolutional neural networks trained on object recognition derive some inspiration from the neuroscience 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 hierarchical organization of primates, the visual system of the mouse has flatter hierarchy. Since mice are capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a framework for building a biologically constrained convolutional neural network model of lateral areas of the mouse visual cortex. The structural parameters of the network are derived from experimental measurements, specifically estimates of numbers of neurons in each area and cortical layer, the interareal connec-tome, 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. The code is freely available to support such research.