Modular representations emerge in neural networks trained to perform context-dependent tasks.

The brain has large-scale modular structure in the form of brain regions, which are thought to arise from constraints on connectivity and the physical geometry of the cortical sheet. In contrast, experimental and theoretical work has argued both for and against the existence of specialized sub-populations of neurons (modules) within single brain regions. By studying artificial neural networks, we show that this local modularity emerges to support context-dependent behavior, but only when the input is low-dimensional. No anatomical constraints are required. We also show when modular specialization emerges at the population level (different modules correspond to orthogonal subspaces). Modularity yields abstract representations, allows for rapid learning and generalization on novel tasks, and facilitates the rapid learning of related contexts. Non-modular representations facilitate the rapid learning of unrelated contexts. Our findings reconcile conflicting experimental results and make predictions for future experiments.

Purkinje cells in Crus I and II encode the visual stimulus and the impending choice as monkeys learn a reinforcement based visuomotor association task.

Visuomotor association involves linking an arbitrary visual cue to a well-learned movement. Transient inactivation of Crus I/II impairs primates' ability to learn new associations and delays motor responses without affecting the kinematics of the movement. The simple spikes of Purkinje cells in the Crus regions signal cognitive errors as monkeys learn to associate specific fractal stimuli with movements of the left or right hand. Here we show that as learning progresses, the simple spike activity of individual neurons becomes more selective for stimulus-response associations, with selectivity developing closer to the appearance of visual stimuli. Initially, most neurons respond to both associations, irrespective of the identity of the stimulus and the associated movement, but as learning advances, more neurons distinguish between specific stimulus-hand associations. Using a linear decoder, it was found that in early learning stages, the visual stimulus can be decoded only when the choice can also be decoded. As learning improves, the visual stimulus is decoded earlier than the choice. A simple model can replicate the observed simple spike signals and the monkeys' behavior in both the early and late learning stages.

Abstract representations emerge in human hippocampal neurons during inference.

Humans have the remarkable cognitive capacity to rapidly adapt to changing environments. Central to this capacity is the ability to form high-level, abstract representations that take advantage of regularities in the world to support generalization1. However, little is known about how these representations are encoded in populations of neurons, how they emerge through learning and how they relate to behaviour2,3. Here we characterized the representational geometry of populations of neurons (single units) recorded in the hippocampus, amygdala, medial frontal cortex and ventral temporal cortex of neurosurgical patients performing an inferential reasoning task. We found that only the neural representations formed in the hippocampus simultaneously encode several task variables in an abstract, or disentangled, format. This representational geometry is uniquely observed after patients learn to perform inference, and consists of disentangled directly observable and discovered latent task variables. Learning to perform inference by trial and error or through verbal instructions led to the formation of hippocampal representations with similar geometric properties. The observed relation between representational format and inference behaviour suggests that abstract and disentangled representational geometries are important for complex cognition.

Neural representational geometries reflect behavioral differences in monkeys and recurrent neural networks.

Animals likely use a variety of strategies to solve laboratory tasks. Traditionally, combined analysis of behavioral and neural recording data across subjects employing different strategies may obscure important signals and give confusing results. Hence, it is essential to develop techniques that can infer strategy at the single-subject level. We analyzed an experiment in which two male monkeys performed a visually cued rule-based task. The analysis of their performance shows no indication that they used a different strategy. However, when we examined the geometry of stimulus representations in the state space of the neural activities recorded in dorsolateral prefrontal cortex, we found striking differences between the two monkeys. Our purely neural results induced us to reanalyze the behavior. The new analysis showed that the differences in representational geometry are associated with differences in the reaction times, revealing behavioral differences we were unaware of. All these analyses suggest that the monkeys are using different strategies. Finally, using recurrent neural network models trained to perform the same task, we show that these strategies correlate with the amount of training, suggesting a possible explanation for the observed neural and behavioral differences.

Perirhinal cortex learns a predictive map of the task environment.

Goal-directed tasks involve acquiring an internal model, known as a predictive map, of relevant stimuli and associated outcomes to guide behavior. Here, we identified neural signatures of a predictive map of task behavior in perirhinal cortex (Prh). Mice learned to perform a tactile working memory task by classifying sequential whisker stimuli over multiple training stages. Chronic two-photon calcium imaging, population analysis, and computational modeling revealed that Prh encodes stimulus features as sensory prediction errors. Prh forms stable stimulus-outcome associations that can progressively be decoded earlier in the trial as training advances and that generalize as animals learn new contingencies. Stimulus-outcome associations are linked to prospective network activity encoding possible expected outcomes. This link is mediated by cholinergic signaling to guide task performance, demonstrated by acetylcholine imaging and systemic pharmacological perturbation. We propose that Prh combines error-driven and map-like properties to acquire a predictive map of learned task behavior.

Temporal multiplexing of perception and memory codes in IT cortex.

A central assumption of neuroscience is that long-term memories are represented by the same brain areas that encode sensory stimuli1. Neurons in inferotemporal (IT) cortex represent the sensory percept of visual objects using a distributed axis code2-4. Whether and how the same IT neural population represents the long-term memory of visual objects remains unclear. Here we examined how familiar faces are encoded in the IT anterior medial face patch (AM), perirhinal face patch (PR) and temporal pole face patch (TP). In AM and PR we observed that the encoding axis for familiar faces is rotated relative to that for unfamiliar faces at long latency; in TP this memory-related rotation was much weaker. Contrary to previous claims, the relative response magnitude to familiar versus unfamiliar faces was not a stable indicator of familiarity in any patch5-11. The mechanism underlying the memory-related axis change is likely intrinsic to IT cortex, because inactivation of PR did not affect axis change dynamics in AM. Overall, our results suggest that memories of familiar faces are represented in AM and perirhinal cortex by a distinct long-latency code, explaining how the same cell population can encode both the percept and memory of faces.

Mixed selectivity: Cellular computations for complexity.

The property of mixed selectivity has been discussed at a computational level and offers a strategy to maximize computational power by adding versatility to the functional role of each neuron. Here, we offer a biologically grounded implementational-level mechanistic explanation for mixed selectivity in neural circuits. We define pure, linear, and nonlinear mixed selectivity and discuss how these response properties can be obtained in simple neural circuits. Neurons that respond to multiple, statistically independent variables display mixed selectivity. If their activity can be expressed as a weighted sum, then they exhibit linear mixed selectivity; otherwise, they exhibit nonlinear mixed selectivity. Neural representations based on diverse nonlinear mixed selectivity are high dimensional; hence, they confer enormous flexibility to a simple downstream readout neural circuit. However, a simple neural circuit cannot possibly encode all possible mixtures of variables simultaneously, as this would require a combinatorially large number of mixed selectivity neurons. Gating mechanisms like oscillations and neuromodulation can solve this problem by dynamically selecting which variables are mixed and transmitted to the readout.

Computational role of structure in neural activity and connectivity.

One major challenge of neuroscience is identifying structure in seemingly disorganized neural activity. Different types of structure have different computational implications that can help neuroscientists understand the functional role of a particular brain area. Here, we outline a unified approach to characterize structure by inspecting the representational geometry and the modularity properties of the recorded activity and show that a similar approach can also reveal structure in connectivity. We start by setting up a general framework for determining geometry and modularity in activity and connectivity and relating these properties with computations performed by the network. We then use this framework to review the types of structure found in recent studies of model networks performing three classes of computations.

Tuned geometries of hippocampal representations meet the computational demands of social memory.

Social memory consists of two processes: the detection of familiar compared with novel conspecifics and the detailed recollection of past social episodes. We investigated the neural bases for these processes using calcium imaging of dorsal CA2 hippocampal pyramidal neurons, known to be important for social memory, during social/spatial encounters with novel conspecifics and familiar littermates. Whereas novel individuals were represented in a low-dimensional geometry that allows for generalization of social identity across different spatial locations and of location across different identities, littermates were represented in a higher-dimensional geometry that supports high-capacity memory storage. Moreover, familiarity was represented in an abstract format, independent of individual identity. The degree to which familiarity increased the dimensionality of CA2 representations for individual mice predicted their performance in a social novelty recognition memory test. Thus, by tuning the geometry of structured neural activity, CA2 is able to meet the demands of distinct social memory processes.

Identifying and modulating neural signatures of stress susceptibility and resilience enables control of anhedonia.

Anhedonia is a core aspect of major depressive disorder. Traditionally viewed as a blunted emotional state in which individuals are unable to experience joy, anhedonia also diminishes the drive to seek rewards and the ability to value and learn about them 1-4.The neural underpinnings of anhedonia and how this emotional state drives related behavioral changes remain unclear. Here, we investigated these questions by taking advantage of the fact that when mice are exposed to traumatic social stress, susceptible animals become socially withdrawn and anhedonic, where they cease to seek high-value rewards, while others remain resilient. By performing high density electrophysiological recordings and comparing neural activity patterns of these groups in the basolateral amygdala (BLA) and ventral CA1 (vCA1) of awake behaving animals, we identified neural signatures of susceptibility and resilience to anhedonia. When animals actively sought rewards, BLA activity in resilient mice showed stronger discrimination between upcoming reward choices. In contrast, susceptible mice displayed a rumination-like signature, where BLA neurons encoded the intention to switch or stay on a previously chosen reward. When animals were at rest, the spontaneous BLA activity of susceptible mice was higher dimensional than in controls, reflecting a greater number of distinct neural population states. Notably, this spontaneous activity allowed us to decode group identity and to infer if a mouse had a history of stress better than behavioral outcomes alone. Finally, targeted manipulation of vCA1 inputs to the BLA in susceptible mice rescued dysfunctional neural dynamics, amplified dynamics associated with resilience, and reversed their anhedonic behavior. This work reveals population-level neural signatures that explain individual differences in responses to traumatic stress, and suggests that modulating vCA1-BLA inputs can enhance resilience by regulating these dynamics.

Neural signatures of stress susceptibility and resilience in the amygdala-hippocampal network.

The neural dynamics that underlie divergent anhedonic responses to stress remain unclear. Here, we identified neuronal dynamics in an amygdala-hippocampal circuit that distinguish stress resilience and susceptibility. In a reward-choice task, basolateral amygdala (BLA) activity in resilient mice showed enhanced discrimination of upcoming reward choices. In contrast, a rumination-like signature emerged in the BLA of susceptible mice; a linear decoder could classify the intention to switch or stay on a previously chosen reward. Spontaneous activity in the BLA of susceptible mice was higher dimensional than controls, reflecting the exploration of a larger number of distinct neural states. Manipulation of vCA1-BLA inputs rescued dysfunctional neural dynamics and anhedonia in susceptible mice, suggesting that targeting this pathway can enhance BLA circuit function and ameliorate of depression-related behaviors.

Abstract representations emerge in human hippocampal neurons during inference behavior.

Humans have the remarkable cognitive capacity to rapidly adapt to changing environments. Central to this capacity is the ability to form high-level, abstract representations that take advantage of regularities in the world to support generalization 1 . However, little is known about how these representations are encoded in populations of neurons, how they emerge through learning, and how they relate to behavior 2,3 . Here we characterized the representational geometry of populations of neurons (single-units) recorded in the hippocampus, amygdala, medial frontal cortex, and ventral temporal cortex of neurosurgical patients who are performing an inferential reasoning task. We find that only the neural representations formed in the hippocampus simultaneously encode multiple task variables in an abstract, or disentangled, format. This representational geometry is uniquely observed after patients learn to perform inference, and consisted of disentangled directly observable and discovered latent task variables. Interestingly, learning to perform inference by trial and error or through verbal instructions led to the formation of hippocampal representations with similar geometric properties. The observed relation between representational format and inference behavior suggests that abstract/disentangled representational geometries are important for complex cognition.

Ventral CA1 Population Codes for Anxiety.

The ventral hippocampus is a critical node in the distributed brain network that controls anxiety. Using miniature microscopy and calcium imaging, we recorded ventral CA1 (vCA1) neurons in freely moving mice as they explored variants of classic behavioral assays for anxiety. Unsupervised behavioral segmentation revealed clusters of behavioral motifs that corresponded to exploratory and vigilance-like states. We discovered multiple vCA1 population codes that represented the anxiogenic features of the environment, such as bright light and openness, as well as the moment-to-moment anxiety state of the animals. These population codes possessed distinct generalization properties: neural representations of anxiogenic features were different for open field and elevated plus/zero maze tasks, while neural representations of moment-to-moment anxiety state were similar across both experimental contexts. Our results suggest that anxiety is not tied to the aversive compartments of these mazes but is rather defined by a behavioral state and its corresponding population code that generalizes across environments.

The representational geometry of emotional states in basolateral amygdala.

Sensory stimuli associated with aversive outcomes can cause multiple behavioral responses related to an animal's evolving emotional state. We employed chemogenetic inactivation and two-photon imaging to reveal how the basolateral amygdala (BLA) mediates these state changes. Mice were presented stimuli in a virtual burrow, causing two responses reflecting fear and flight to safety: tremble and ingress into the burrow. Inactivation eliminated differential tremble and ingress to aversive and neutral stimuli without eliminating responses themselves. Multiple variables, including stimulus valence and identity, and being in the tremble or ingressed state, typically modulated each neuron's activity (mixed-selectivity). BLA neural ensembles represented these variables even after neurons with apparent specialized selectivity were eliminated from analyses. Thus, implementing different readouts of BLA ensembles comprised of mixed-selectivity neurons can identify distinct emotional states defined by responses, like tremble for fear and ingress for safety. This mechanism relies on BLA's representational geometry, not its circuit specialization.

The computational role of structure in neural activity and connectivity.

One major challenge of neuroscience is finding interesting structures in a seemingly disorganized neural activity. Often these structures have computational implications that help to understand the functional role of a particular brain area. Here we outline a unified approach to characterize these structures by inspecting the representational geometry and the modularity properties of the recorded activity, and show that this approach can also reveal structures in connectivity. We start by setting up a general framework for determining geometry and modularity in activity and connectivity and relating these properties with computations performed by the network. We then use this framework to review the types of structure found in recent works on model networks performing three classes of computations.

PERIRHINAL CORTEX LEARNS A PREDICTIVE MAP (INTERNAL MODEL) OF THE TASK ENVIRONMENT.

Goal-directed tasks involve acquiring an internal model, known as a predictive map, of relevant stimuli and associated outcomes to guide behavior. Here, we identified neural signatures of a predictive map of task behavior in perirhinal cortex (Prh). Mice learned to perform a tactile working memory task by classifying sequential whisker stimuli over multiple training stages. Chemogenetic inactivation demonstrated that Prh is involved in task learning. Chronic two-photon calcium imaging, population analysis, and computational modeling revealed that Prh encodes stimulus features as sensory prediction errors. Prh forms stable stimulus-outcome associations that expand in a retrospective manner and generalize as animals learn new contingencies. Stimulus-outcome associations are linked to prospective network activity encoding possible expected outcomes. This link is mediated by cholinergic signaling to guide task performance, demonstrated by acetylcholine imaging and perturbation. We propose that Prh combines error-driven and map-like properties to acquire a predictive map of learned task behavior.

Abstract representations emerge naturally in neural networks trained to perform multiple tasks.

Humans and other animals demonstrate a remarkable ability to generalize knowledge across distinct contexts and objects during natural behavior. We posit that this ability to generalize arises from a specific representational geometry, that we call abstract and that is referred to as disentangled in machine learning. These abstract representations have been observed in recent neurophysiological studies. However, it is unknown how they emerge. Here, using feedforward neural networks, we demonstrate that the learning of multiple tasks causes abstract representations to emerge, using both supervised and reinforcement learning. We show that these abstract representations enable few-sample learning and reliable generalization on novel tasks. We conclude that abstract representations of sensory and cognitive variables may emerge from the multiple behaviors that animals exhibit in the natural world, and, as a consequence, could be pervasive in high-level brain regions. We also make several specific predictions about which variables will be represented abstractly.

Face familiarity detection with complex synapses.

Synaptic plasticity is a complex phenomenon involving multiple biochemical processes that operate on different timescales. Complexity can greatly increase memory capacity when the variables characterizing the synaptic dynamics have limited precision, as shown in simple memory retrieval problems involving random patterns. Here we turn to a real-world problem, face familiarity detection, and we show that synaptic complexity can be harnessed to store in memory a large number of faces that can be recognized at a later time. The number of recognizable faces grows almost linearly with the number of synapses and quadratically with the number of neurons. Complex synapses outperform simple ones characterized by a single variable, even when the total number of dynamical variables is matched. Complex and simple synapses have distinct signatures that are testable in experiments. Our results indicate that a system with complex synapses can be used in real-world tasks such as face familiarity detection.

The geometry of cortical representations of touch in rodents.

Neurons often encode highly heterogeneous non-linear functions of multiple task variables, a signature of a high-dimensional geometry. We studied the representational geometry in the somatosensory cortex of mice trained to report the curvature of objects touched by their whiskers. High-speed videos of the whiskers revealed that the task can be solved by linearly integrating multiple whisker contacts over time. However, the neural activity in somatosensory cortex reflects non-linear integration of spatio-temporal features of the sensory inputs. Although the responses at first appeared disorganized, we identified an interesting structure in the representational geometry: different whisker contacts are disentangled variables represented in approximately, but not fully, orthogonal subspaces of the neural activity space. This geometry allows linear readouts to perform a broad class of tasks of different complexities without compromising the ability to generalize to novel situations.