Constructing neural network models from brain data reveals representational transformations linked to adaptive behavior

Flexible gating between subspaces in a neural network model of internally guided task switching

Behavioral flexibility relies on the brain's ability to switch rapidly between multiple tasks, even when the task rule is not explicitly cued but must be inferred through trial and error. The underlying neural circuit mechanism remains poorly understood. We investigated recurrent neural networks (RNNs) trained to perform an analog of the classic Wisconsin Card Sorting Test. The networks consist of two modules responsible for rule representation and sensorimotor mapping, respectively, where each module is comprised of a circuit with excitatory neurons and three major types of inhibitory neurons. We found that rule representation by self-sustained persistent activity across trials, error monitoring and gated sensorimotor mapping emerged from training. Systematic dissection of trained RNNs revealed a detailed circuit mechanism that is consistent across networks trained with different hyperparameters. The networks' dynamical trajectories for different rules resided in separate subspaces of population activity; the subspaces collapsed and performance was reduced to chance level when dendrite-targeting somatostatin-expressing interneurons were silenced, illustrating how a phenomenological description of representational subspaces is explained by a specific circuit mechanism.

Emergence and reconfiguration of modular structure for artificial neural networks during continual familiarity detection

Advances in artificial intelligence enable neural networks to learn a wide variety of tasks, yet our understanding of the learning dynamics of these networks remains limited. Here, we study the temporal dynamics during learning of Hebbian feedforward neural networks in tasks of continual familiarity detection. Drawing inspiration from network neuroscience, we examine the network's dynamic reconfiguration, focusing on how network modules evolve throughout learning. Through a comprehensive assessment involving metrics like network accuracy, modular flexibility, and distribution entropy across diverse learning modes, our approach reveals various previously unknown patterns of network reconfiguration. We find that the emergence of network modularity is a salient predictor of performance and that modularization strengthens with increasing flexibility throughout learning. These insights not only elucidate the nuanced interplay of network modularity, accuracy, and learning dynamics but also bridge our understanding of learning in artificial and biological agents.

Timescales of learning in prefrontal cortex

The lateral prefrontal cortex (PFC) in humans and other primates is critical for immediate, goal-directed behaviour and working memory, which are classically considered distinct from the cognitive and neural circuits that support long-term learning and memory. Over the past few years, a reconsideration of this textbook perspective has emerged, in that different timescales of memory-guided behaviour are in constant interaction during the pursuit of immediate goals. Here, we will first detail how neural activity related to the shortest timescales of goal-directed behaviour (which requires maintenance of current states and goals in working memory) is sculpted by long-term knowledge and learning - that is, how the past informs present behaviour. Then, we will outline how learning across different timescales (from seconds to years) drives plasticity in the primate lateral PFC, from single neuron firing rates to mesoscale neuroimaging activity patterns. Finally, we will review how, over days and months of learning, dense local and long-range connectivity patterns in PFC facilitate longer-lasting changes in population activity by changing synaptic weights and recruiting additional neural resources to inform future behaviour. Our Review sheds light on how the machinery of plasticity in PFC circuits facilitates the integration of learned experiences across time to best guide adaptive behaviour.

Flexible gating between subspaces in a neural network model of internally guided task switching

Behavioral flexibility relies on the brain's ability to switch rapidly between multiple tasks, even when the task rule is not explicitly cued but must be inferred through trial and error. The underlying neural circuit mechanism remains poorly understood. We investigated recurrent neural networks (RNNs) trained to perform an analog of the classic Wisconsin Card Sorting Test. The networks consist of two modules responsible for rule representation and sensorimotor mapping, respectively, where each module is comprised of a circuit with excitatory neurons and three major types of inhibitory neurons. We found that rule representation by self-sustained persistent activity across trials, error monitoring and gated sensorimotor mapping emerged from training. Systematic dissection of trained RNNs revealed a detailed circuit mechanism that is consistent across networks trained with different hyperparameters. The networks' dynamical trajectories for different rules resided in separate subspaces of population activity; the subspaces collapsed and performance was reduced to chance level when dendrite-targeting somatostatin-expressing interneurons were silenced, illustrating how a phenomenological description of representational subspaces is explained by a specific circuit mechanism.

The Prediction of Brain Activity from Connectivity: Advances and Applications

The human brain is composed of multiple, discrete, functionally specialized regions that are interconnected to form large-scale distributed networks. Using advanced brain-imaging methods and machine-learning analytical approaches, recent studies have demonstrated that regional brain activity during the performance of various cognitive tasks can be accurately predicted from patterns of task-independent brain connectivity. In this review article, we first present evidence for the predictability of brain activity from structural connectivity (i.e., white matter connections) and functional connectivity (i.e., temporally synchronized task-free activations). We then discuss the implications of such predictions to clinical populations, such as patients diagnosed with psychiatric disorders or neurologic diseases, and to the study of brain-behavior associations. We conclude that connectivity may serve as an infrastructure that dictates brain activity, and we pinpoint several open questions and directions for future research.

Natural language instructions induce compositional generalization in networks of neurons

A fundamental human cognitive feat is to interpret linguistic instructions in order to perform novel tasks without explicit task experience. Yet, the neural computations that might be used to accomplish this remain poorly understood. We use advances in natural language processing to create a neural model of generalization based on linguistic instructions. Models are trained on a set of common psychophysical tasks, and receive instructions embedded by a pretrained language model. Our best models can perform a previously unseen task with an average performance of 83% correct based solely on linguistic instructions (that is, zero-shot learning). We found that language scaffolds sensorimotor representations such that activity for interrelated tasks shares a common geometry with the semantic representations of instructions, allowing language to cue the proper composition of practiced skills in unseen settings. We show how this model generates a linguistic description of a novel task it has identified using only motor feedback, which can subsequently guide a partner model to perform the task. Our models offer several experimentally testable predictions outlining how linguistic information must be represented to facilitate flexible and general cognition in the human brain.

Machine learning and artificial intelligence in neuroscience: A primer for researchers

Artificial intelligence (AI) is often used to describe the automation of complex tasks that we would attribute intelligence to. Machine learning (ML) is commonly understood as a set of methods used to develop an AI. Both have seen a recent boom in usage, both in scientific and commercial fields. For the scientific community, ML can solve bottle necks created by complex, multi-dimensional data generated, for example, by functional brain imaging or *omics approaches. ML can here identify patterns that could not have been found using traditional statistic approaches. However, ML comes with serious limitations that need to be kept in mind: their tendency to optimise solutions for the input data means it is of crucial importance to externally validate any findings before considering them more than a hypothesis. Their black-box nature implies that their decisions usually cannot be understood, which renders their use in medical decision making problematic and can lead to ethical issues. Here, we present an introduction for the curious to the field of ML/AI. We explain the principles as commonly used methods as well as recent methodological advancements before we discuss risks and what we see as future directions of the field. Finally, we show practical examples of neuroscience to illustrate the use and limitations of ML.

Functional neuroimaging as a catalyst for integrated neuroscience

Functional magnetic resonance imaging (fMRI) enables non-invasive access to the awake, behaving human brain. By tracking whole-brain signals across a diverse range of cognitive and behavioural states or mapping differences associated with specific traits or clinical conditions, fMRI has advanced our understanding of brain function and its links to both normal and atypical behaviour. Despite this headway, progress in human cognitive neuroscience that uses fMRI has been relatively isolated from rapid advances in other subdomains of neuroscience, which themselves are also somewhat siloed from one another. In this Perspective, we argue that fMRI is well-placed to integrate the diverse subfields of systems, cognitive, computational and clinical neuroscience. We first summarize the strengths and weaknesses of fMRI as an imaging tool, then highlight examples of studies that have successfully used fMRI in each subdomain of neuroscience. We then provide a roadmap for the future advances that will be needed to realize this integrative vision. In this way, we hope to demonstrate how fMRI can help usher in a new era of interdisciplinary coherence in neuroscience.

Evaluating functional brain organization in individuals and identifying contributions to network overlap

Individual differences in the spatial organization of resting state networks have received increased attention in recent years. Measures of individual-specific spatial organization of brain networks and overlapping network organization have been linked to important behavioral and clinical traits and are therefore potential biomarker targets for personalized psychiatry approaches. To better understand individual-specific spatial brain organization, this paper addressed three key goals. First, we determined whether it is possible to reliably estimate weighted (non-binarized) resting state network maps using data from only a single individual, while also maintaining maximum spatial correspondence across individuals. Second, we determined the degree of spatial overlap between distinct networks, using test-retest and twin data. Third, we systematically tested multiple hypotheses (spatial mixing, temporal switching, and coupling) as candidate explanations for why networks overlap spatially. To estimate weighted network organization, we adopt the Probabilistic Functional Modes (PROFUMO) algorithm, which implements a Bayesian framework with hemodynamic and connectivity priors to supplement optimization for spatial sparsity/independence. Our findings showed that replicable individual-specific estimates of weighted resting state networks can be derived using high quality fMRI data within individual subjects. Network organization estimates using only data from each individual subject closely resembled group-informed network estimates (which was not explicitly modeled in our individual-specific analyses), suggesting that cross-subject correspondence was largely maintained. Furthermore, our results confirmed the presence of spatial overlap in network organization, which was replicable across sessions within individuals and in monozygotic twin pairs. Intriguingly, our findings provide evidence that network overlap is indicative of linear additive coupling. These results suggest that regions of network overlap concurrently process information from all contributing networks, potentially pointing to the role of overlapping network organization in the integration of information across multiple brain systems.

Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations

Brain activity flow models estimate the movement of task-evoked activity over brain connections to help explain network-generated task functionality. Activity flow models have been shown to accurately generate task-evoked brain activations across a wide variety of brain regions and task conditions. However, these models have had limited explanatory power, given known issues with causal interpretations of the standard functional connectivity measures used to parameterize activity flow models. We show here that functional/effective connectivity (FC) measures grounded in causal principles facilitate mechanistic interpretation of activity flow models. We progress from simple to complex FC measures, with each adding algorithmic details reflecting causal principles. This reflects many neuroscientists' preference for reduced FC measure complexity (to minimize assumptions, minimize compute time, and fully comprehend and easily communicate methodological details), which potentially trades off with causal validity. We start with Pearson correlation (the current field standard) to remain maximally relevant to the field, estimating causal validity across a range of FC measures using simulations and empirical fMRI data. Finally, we apply causal-FC-based activity flow modeling to a dorsolateral prefrontal cortex region (DLPFC), demonstrating distributed causal network mechanisms contributing to its strong activation during a working memory task. Notably, this fully distributed model is able to account for DLPFC working memory effects traditionally thought to rely primarily on within-region (i.e., not distributed) recurrent processes. Together, these results reveal the promise of parameterizing activity flow models using causal FC methods to identify network mechanisms underlying cognitive computations in the human brain.

Network models to enhance the translational impact of cross-species studies

Neuroscience studies are often carried out in animal models for the purpose of understanding specific aspects of the human condition. However, the translation of findings across species remains a substantial challenge. Network science approaches can enhance the translational impact of cross-species studies by providing a means of mapping small-scale cellular processes identified in animal model studies to larger-scale inter-regional circuits observed in humans. In this Review, we highlight the contributions of network science approaches to the development of cross-species translational research in neuroscience. We lay the foundation for our discussion by exploring the objectives of cross-species translational models. We then discuss how the development of new tools that enable the acquisition of whole-brain data in animal models with cellular resolution provides unprecedented opportunity for cross-species applications of network science approaches for understanding large-scale brain networks. We describe how these tools may support the translation of findings across species and imaging modalities and highlight future opportunities. Our overarching goal is to illustrate how the application of network science tools across human and animal model studies could deepen insight into the neurobiology that underlies phenomena observed with non-invasive neuroimaging methods and could simultaneously further our ability to translate findings across species.

Neural representation dynamics reveal computational principles of cognitive task learning

During cognitive task learning, neural representations must be rapidly constructed for novel task performance, then optimized for robust practiced task performance. How the geometry of neural representations changes to enable this transition from novel to practiced performance remains unknown. We hypothesized that practice involves a shift from compositional representations (task-general activity patterns that can be flexibly reused across tasks) to conjunctive representations (task-specific activity patterns specialized for the current task). Functional MRI during learning of multiple complex tasks substantiated this dynamic shift from compositional to conjunctive representations, which was associated with reduced cross-task interference (via pattern separation) and behavioral improvement. Further, we found that conjunctions originated in subcortex (hippocampus and cerebellum) and slowly spread to cortex, extending multiple memory systems theories to encompass task representation learning. The formation of conjunctive representations hence serves as a computational signature of learning, reflecting cortical-subcortical dynamics that optimize task representations in the human brain.

A Transient High-dimensional Geometry Affords Stable Conjunctive Subspaces for Efficient Action Selection

Flexible action selection requires cognitive control mechanisms capable of mapping the same inputs to diverse output actions depending on goals and contexts. How the brain encodes information to enable this capacity remains one of the longstanding and fundamental problems in cognitive neuroscience. From a neural state-space perspective, solving this problem requires a control representation that can disambiguate similar input neural states, making task-critical dimensions separable depending on the context. Moreover, for action selection to be robust and time-invariant, control representations must be stable in time, thereby enabling efficient readout by downstream processing units. Thus, an ideal control representation should leverage geometry and dynamics that maximize the separability and stability of neural trajectories for task computations. Here, using novel EEG decoding methods, we investigated how the geometry and dynamics of control representations constrain flexible action selection in the human brain. Specifically, we tested the hypothesis that encoding a temporally stable conjunctive subspace that integrates stimulus, response, and context (i.e., rule) information in a high-dimensional geometry achieves the separability and stability needed for context-dependent action selection. Human participants performed a task that requires context-dependent action selection based on pre-instructed rules. Participants were cued to respond immediately at varying intervals following stimulus presentation, which forced responses at different states in neural trajectories. We discovered that in the moments before successful responses, there was a transient expansion of representational dimensionality that separated conjunctive subspaces. Further, we found that the dynamics stabilized in the same time window, and that the timing of entry into this stable and high-dimensional state predicted the quality of response selection on individual trials. These results establish the neural geometry and dynamics the human brain needs for flexible control over behavior.

Flexible sensory-motor mapping rules manifest in correlated variability of stimulus and action codes across the brain

Humans and non-human primates can flexibly switch between different arbitrary mappings from sensation to action to solve a cognitive task. It has remained unknown how the brain implements such flexible sensory-motor mapping rules. Here, we uncovered a dynamic reconfiguration of task-specific correlated variability between sensory and motor brain regions. Human participants switched between two rules for reporting visual orientation judgments during fMRI recordings. Rule switches were either signaled explicitly or inferred by the participants from ambiguous cues. We used behavioral modeling to reconstruct the time course of their belief about the active rule. In both contexts, the patterns of correlations between ongoing fluctuations in stimulus- and action-selective activity across visual- and action-related brain regions tracked participants' belief about the active rule. The rule-specific correlation patterns broke down around the time of behavioral errors. We conclude that internal beliefs about task state are instantiated in brain-wide, selective patterns of correlated variability.

Abnormal brain network community structure related to psychological stress in schizophrenia

Recent functional imaging studies in schizophrenia consistently report a disruption of brain connectivity. However, most of these studies analyze the brain connectivity during resting state. Since psychological stress is a major factor for the emergence of psychotic symptoms, we sought to characterize the brain connectivity reconfiguration induced by stress in schizophrenia. We tested the hypothesis that an alteration of the brain's integration-segregation dynamic could be the result of patients with schizophrenia facing psychological stress. To this end, we studied the modular organization and the reconfiguration of networks induced by a stress paradigm in forty subjects (twenty patients and twenty controls), thus analyzing the dynamics of the brain in terms of integration and segregation processes by using 3T-fMRI. Patients with schizophrenia did not show statistically significant differences during the control task compared with controls, but they showed an abnormal community structure during stress condition and an under-connected reconfiguration network with a reduction of hub nodes, suggesting a deficit of integration dynamic with a greater compromise of the right hemisphere. These results provide evidence that schizophrenia has a normal response to undemanding stimuli but shows a disruption of brain functional connectivity between key regions involved in stress response, potentially leading to altered functional brain dynamics by reducing integration capacity and showing deficits recruiting right hemisphere regions. This could in turn underlie the hyper-sensitivity to stress characteristic of schizophrenia.

Canonical template tracking: Measuring the activation state of specific neural representations

Multivariate analyses of neural data have become increasingly influential in cognitive neuroscience since they allow to address questions about the representational signatures of neurocognitive phenomena. Here, we describe Canonical Template Tracking: a multivariate approach that employs independent localizer tasks to assess the activation state of specific representations during the execution of cognitive paradigms. We illustrate the benefits of this methodology in characterizing the particular content and format of task-induced representations, comparing it with standard (cross-)decoding and representational similarity analyses. Then, we discuss relevant design decisions for experiments using this analysis approach, focusing on the nature of the localizer tasks from which the canonical templates are derived. We further provide a step-by-step tutorial of this method, stressing the relevant analysis choices for functional magnetic resonance imaging and magneto/electroencephalography data. Importantly, we point out the potential pitfalls linked to canonical template tracking implementation and interpretation of the results, together with recommendations to mitigate them. To conclude, we provide some examples from previous literature that highlight the potential of this analysis to address relevant theoretical questions in cognitive neuroscience.

Thalamocortical contributions to cognitive task activity

Thalamocortical interaction is a ubiquitous functional motif in the mammalian brain. Previously (Hwang et al., 2021), we reported that lesions to network hubs in the human thalamus are associated with multi-domain behavioral impairments in language, memory, and executive functions. Here, we show how task-evoked thalamic activity is organized to support these broad cognitive abilities. We analyzed functional magnetic resonance imaging (MRI) data from human subjects that performed 127 tasks encompassing a broad range of cognitive representations. We first investigated the spatial organization of task-evoked activity and found a basis set of activity patterns evoked to support processing needs of each task. Specifically, the anterior, medial, and posterior-medial thalamus exhibit hub-like activity profiles that are suggestive of broad functional participation. These thalamic task hubs overlapped with network hubs interlinking cortical systems. To further determine the cognitive relevance of thalamic activity and thalamocortical functional connectivity, we built a data-driven thalamocortical model to test whether thalamic activity can be used to predict cortical task activity. The thalamocortical model predicted task-specific cortical activity patterns, and outperformed comparison models built on cortical, hippocampal, and striatal regions. Simulated lesions to low-dimensional, multi-task thalamic hub regions impaired task activity prediction. This simulation result was further supported by profiles of neuropsychological impairments in human patients with focal thalamic lesions. In summary, our results suggest a general organizational principle of how the human thalamocortical system supports cognitive task activity.

What is episodic memory and how do we use it?

What are the neural and computational principles that give rise to episodic memory? Although memory is probably the most studied topic in psychology and cognitive neuroscience, most research has focused on learning at the micro-level. I outline the limitations of this approach and propose a 'molar' approach to tackle episodic memory at the scale of life.