Computational Mechanisms of Predictive Processing in Brains and Machines

Predictive processing offers a unifying view of neural computation, proposing that brains continuously anticipate sensory input and update internal models based on prediction errors. In this talk, I will present converging evidence for the computational mechanisms underlying this framework across human neuroscience and deep neural networks. I will begin with recent work showing that large-scale distributed prediction-error encoding in the human brain directly predicts how sensory representations reorganize through predictive learning. I will then turn to PredNet, a popular predictive coding inspired deep network that has been widely used to model real-world biological vision systems. Using dynamic stimuli generated with our Spatiotemporal Style Transfer algorithm, we demonstrate that PredNet relies primarily on low-level spatiotemporal structure and remains insensitive to high-level content, revealing limits in its generalization capacity. Finally, I will discuss new recurrent vision models that integrate top-down feedback connections with intrinsic neural variability, uncovering a dual mechanism for robust sensory coding in which neural variability decorrelates unit responses, while top-down feedback stabilizes network dynamics. Together, these results outline how prediction error signaling and top-down feedback pathways shape adaptive sensory processing in biological and artificial systems.

Neural Representations of Abstract Cognitive Maps in Prefrontal Cortex and Medial Temporal Lobe

From Spiking Predictive Coding to Learning Abstract Object Representation

In a first part of the talk, I will present Predictive Coding Light (PCL), a novel unsupervised learning architecture for spiking neural networks. In contrast to conventional predictive coding approaches, which only transmit prediction errors to higher processing stages, PCL learns inhibitory lateral and top-down connectivity to suppress the most predictable spikes and passes a compressed representation of the input to higher processing stages. We show that PCL reproduces a range of biological findings and exhibits a favorable tradeoff between energy consumption and downstream classification performance on challenging benchmarks. A second part of the talk will feature our lab’s efforts to explain how infants and toddlers might learn abstract object representations without supervision. I will present deep learning models that exploit the temporal and multimodal structure of their sensory inputs to learn representations of individual objects, object categories, or abstract super-categories such as „kitchen object“ in a fully unsupervised fashion. These models offer a parsimonious account of how abstract semantic knowledge may be rooted in children's embodied first-person experiences.

Cognitive maps as expectations learned across episodes – a model of the two dentate gyrus blades

How can the hippocampal system transition from episodic one-shot learning to a multi-shot learning regime and what is the utility of the resultant neural representations? This talk will explore the role of the dentate gyrus (DG) anatomy in this context. The canonical DG model suggests it performs pattern separation. More recent experimental results challenge this standard model, suggesting DG function is more complex and also supports the precise binding of objects and events to space and the integration of information across episodes. Very recent studies attribute pattern separation and pattern integration to anatomically distinct parts of the DG (the suprapyramidal blade vs the infrapyramidal blade). We propose a computational model that investigates this distinction. In the model the two processing streams (potentially localized in separate blades) contribute to the storage of distinct episodic memories, and the integration of information across episodes, respectively. The latter forms generalized expectations across episodes, eventually forming a cognitive map. We train the model with two data sets, MNIST and plausible entorhinal cortex inputs. The comparison between the two streams allows for the calculation of a prediction error, which can drive the storage of poorly predicted memories and the forgetting of well-predicted memories. We suggest that differential processing across the DG aids in the iterative construction of spatial cognitive maps to serve the generation of location-dependent expectations, while at the same time preserving episodic memory traces of idiosyncratic events.

Learning Representations of Complex Meaning in the Human Brain

Dimensionality reduction beyond neural subspaces

Over the past decade, neural representations have been studied from the lens of low-dimensional subspaces defined by the co-activation of neurons. However, this view has overlooked other forms of covarying structure in neural activity, including i) condition-specific high-dimensional neural sequences, and ii) representations that change over time due to learning or drift. In this talk, I will present a new framework that extends the classic view towards additional types of covariability that are not constrained to a fixed, low-dimensional subspace. In addition, I will present sliceTCA, a new tensor decomposition that captures and demixes these different types of covariability to reveal task-relevant structure in neural activity. Finally, I will close with some thoughts regarding the circuit mechanisms that could generate mixed covariability. Together this work points to a need to consider new possibilities for how neural populations encode sensory, cognitive, and behavioral variables beyond neural subspaces.

Contentopic mapping and object dimensionality - a novel understanding on the organization of object knowledge

Our ability to recognize an object amongst many others is one of the most important features of the human mind. However, object recognition requires tremendous computational effort, as we need to solve a complex and recursive environment with ease and proficiency. This challenging feat is dependent on the implementation of an effective organization of knowledge in the brain. Here I put forth a novel understanding of how object knowledge is organized in the brain, by proposing that the organization of object knowledge follows key object-related dimensions, analogously to how sensory information is organized in the brain. Moreover, I will also put forth that this knowledge is topographically laid out in the cortical surface according to these object-related dimensions that code for different types of representational content – I call this contentopic mapping. I will show a combination of fMRI and behavioral data to support these hypotheses and present a principled way to explore the multidimensionality of object processing.

Rethinking Attention: Dynamic Prioritization

Decades of research on understanding the mechanisms of attentional selection have focused on identifying the units (representations) on which attention operates in order to guide prioritized sensory processing. These attentional units fit neatly to accommodate our understanding of how attention is allocated in a top-down, bottom-up, or historical fashion. In this talk, I will focus on attentional phenomena that are not easily accommodated within current theories of attentional selection – the “attentional platypuses,” as they allude to an observation that within biological taxonomies the platypus does not fit into either mammal or bird categories. Similarly, attentional phenomena that do not fit neatly within current attentional models suggest that current models need to be revised. I list a few instances of the ‘attentional platypuses” and then offer a new approach, the Dynamically Weighted Prioritization, stipulating that multiple factors impinge onto the attentional priority map, each with a corresponding weight. The interaction between factors and their corresponding weights determines the current state of the priority map which subsequently constrains/guides attention allocation. I propose that this new approach should be considered as a supplement to existing models of attention, especially those that emphasize categorical organizations.

LLMs and Human Language Processing

This webinar convened researchers at the intersection of Artificial Intelligence and Neuroscience to investigate how large language models (LLMs) can serve as valuable “model organisms” for understanding human language processing. Presenters showcased evidence that brain recordings (fMRI, MEG, ECoG) acquired while participants read or listened to unconstrained speech can be predicted by representations extracted from state-of-the-art text- and speech-based LLMs. In particular, text-based LLMs tend to align better with higher-level language regions, capturing more semantic aspects, while speech-based LLMs excel at explaining early auditory cortical responses. However, purely low-level features can drive part of these alignments, complicating interpretations. New methods, including perturbation analyses, highlight which linguistic variables matter for each cortical area and time scale. Further, “brain tuning” of LLMs—fine-tuning on measured neural signals—can improve semantic representations and downstream language tasks. Despite open questions about interpretability and exact neural mechanisms, these results demonstrate that LLMs provide a promising framework for probing the computations underlying human language comprehension and production at multiple spatiotemporal scales.

Brain circuits for spatial navigation

In this webinar on spatial navigation circuits, three researchers—Ann Hermundstad, Ila Fiete, and Barbara Webb—discussed how diverse species solve navigation problems using specialized yet evolutionarily conserved brain structures. Hermundstad illustrated the fruit fly’s central complex, focusing on how hardwired circuit motifs (e.g., sinusoidal steering curves) enable rapid, flexible learning of goal-directed navigation. This framework combines internal heading representations with modifiable goal signals, leveraging activity-dependent plasticity to adapt to new environments. Fiete explored the mammalian head-direction system, demonstrating how population recordings reveal a one-dimensional ring attractor underlying continuous integration of angular velocity. She showed that key theoretical predictions—low-dimensional manifold structure, isometry, uniform stability—are experimentally validated, underscoring parallels to insect circuits. Finally, Webb described honeybee navigation, featuring path integration, vector memories, route optimization, and the famous waggle dance. She proposed that allocentric velocity signals and vector manipulation within the central complex can encode and transmit distances and directions, enabling both sophisticated foraging and inter-bee communication via dance-based cues.

Unmotivated bias

In this talk, I will explore how social affective biases arise even in the absence of motivational factors as an emergent outcome of the basic structure of social learning. In several studies, we found that initial negative interactions with some members of a group can cause subsequent avoidance of the entire group, and that this avoidance perpetuates stereotypes. Additional cognitive modeling discovered that approach and avoidance behavior based on biased beliefs not only influences the evaluative (positive or negative) impressions of group members, but also shapes the depth of the cognitive representations available to learn about individuals. In other words, people have richer cognitive representations of members of groups that are not avoided, akin to individualized vs group level categories. I will end presenting a series of multi-agent reinforcement learning simulations that demonstrate the emergence of these social-structural feedback loops in the development and maintenance of affective biases.

Mapping the Brain‘s Visual Representations Using Deep Learning

Navigating semantic spaces: recycling the brain GPS for higher-level cognition

Humans share with other animals a complex neuronal machinery that evolved to support navigation in the physical space and that supports wayfinding and path integration. In my talk I will present a series of recent neuroimaging studies in humans performed in my Lab aimed at investigating the idea that this same neural navigation system (the “brain GPS”) is also used to organize and navigate concepts and memories, and that abstract and spatial representations rely on a common neural fabric. I will argue that this might represent a novel example of “cortical recycling”, where the neuronal machinery that primarily evolved, in lower level animals, to represent relationships between spatial locations and navigate space, in humans are reused to encode relationships between concepts in an internal abstract representational space of meaning.

Learning representations of specifics and generalities over time

There is a fundamental tension between storing discrete traces of individual experiences, which allows recall of particular moments in our past without interference, and extracting regularities across these experiences, which supports generalization and prediction in similar situations in the future. One influential proposal for how the brain resolves this tension is that it separates the processes anatomically into Complementary Learning Systems, with the hippocampus rapidly encoding individual episodes and the neocortex slowly extracting regularities over days, months, and years. But this does not explain our ability to learn and generalize from new regularities in our environment quickly, often within minutes. We have put forward a neural network model of the hippocampus that suggests that the hippocampus itself may contain complementary learning systems, with one pathway specializing in the rapid learning of regularities and a separate pathway handling the region’s classic episodic memory functions. This proposal has broad implications for how we learn and represent novel information of specific and generalized types, which we test across statistical learning, inference, and category learning paradigms. We also explore how this system interacts with slower-learning neocortical memory systems, with empirical and modeling investigations into how the hippocampus shapes neocortical representations during sleep. Together, the work helps us understand how structured information in our environment is initially encoded and how it then transforms over time.

Learning produces a hippocampal cognitive map in the form of an orthogonalized state machine

Cognitive maps confer animals with flexible intelligence by representing spatial, temporal, and abstract relationships that can be used to shape thought, planning, and behavior. Cognitive maps have been observed in the hippocampus, but their algorithmic form and the processes by which they are learned remain obscure. Here, we employed large-scale, longitudinal two-photon calcium imaging to record activity from thousands of neurons in the CA1 region of the hippocampus while mice learned to efficiently collect rewards from two subtly different versions of linear tracks in virtual reality. The results provide a detailed view of the formation of a cognitive map in the hippocampus. Throughout learning, both the animal behavior and hippocampal neural activity progressed through multiple intermediate stages, gradually revealing improved task representation that mirrored improved behavioral efficiency. The learning process led to progressive decorrelations in initially similar hippocampal neural activity within and across tracks, ultimately resulting in orthogonalized representations resembling a state machine capturing the inherent struture of the task. We show that a Hidden Markov Model (HMM) and a biologically plausible recurrent neural network trained using Hebbian learning can both capture core aspects of the learning dynamics and the orthogonalized representational structure in neural activity. In contrast, we show that gradient-based learning of sequence models such as Long Short-Term Memory networks (LSTMs) and Transformers do not naturally produce such orthogonalized representations. We further demonstrate that mice exhibited adaptive behavior in novel task settings, with neural activity reflecting flexible deployment of the state machine. These findings shed light on the mathematical form of cognitive maps, the learning rules that sculpt them, and the algorithms that promote adaptive behavior in animals. The work thus charts a course toward a deeper understanding of biological intelligence and offers insights toward developing more robust learning algorithms in artificial intelligence.

Unifying the mechanisms of hippocampal episodic memory and prefrontal working memory

Remembering events in the past is crucial to intelligent behaviour. Flexible memory retrieval, beyond simple recall, requires a model of how events relate to one another. Two key brain systems are implicated in this process: the hippocampal episodic memory (EM) system and the prefrontal working memory (WM) system. While an understanding of the hippocampal system, from computation to algorithm and representation, is emerging, less is understood about how the prefrontal WM system can give rise to flexible computations beyond simple memory retrieval, and even less is understood about how the two systems relate to each other. Here we develop a mathematical theory relating the algorithms and representations of EM and WM by showing a duality between storing memories in synapses versus neural activity. In doing so, we develop a formal theory of the algorithm and representation of prefrontal WM as structured, and controllable, neural subspaces (termed activity slots). By building models using this formalism, we elucidate the differences, similarities, and trade-offs between the hippocampal and prefrontal algorithms. Lastly, we show that several prefrontal representations in tasks ranging from list learning to cue dependent recall are unified as controllable activity slots. Our results unify frontal and temporal representations of memory, and offer a new basis for understanding the prefrontal representation of WM

Trends in NeuroAI - Meta's MEG-to-image reconstruction

Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri). Title: Brain-optimized inference improves reconstructions of fMRI brain activity Abstract: The release of large datasets and developments in AI have led to dramatic improvements in decoding methods that reconstruct seen images from human brain activity. We evaluate the prospect of further improving recent decoding methods by optimizing for consistency between reconstructions and brain activity during inference. We sample seed reconstructions from a base decoding method, then iteratively refine these reconstructions using a brain-optimized encoding model that maps images to brain activity. At each iteration, we sample a small library of images from an image distribution (a diffusion model) conditioned on a seed reconstruction from the previous iteration. We select those that best approximate the measured brain activity when passed through our encoding model, and use these images for structural guidance during the generation of the small library in the next iteration. We reduce the stochasticity of the image distribution at each iteration, and stop when a criterion on the "width" of the image distribution is met. We show that when this process is applied to recent decoding methods, it outperforms the base decoding method as measured by human raters, a variety of image feature metrics, and alignment to brain activity. These results demonstrate that reconstruction quality can be significantly improved by explicitly aligning decoding distributions to brain activity distributions, even when the seed reconstruction is output from a state-of-the-art decoding algorithm. Interestingly, the rate of refinement varies systematically across visual cortex, with earlier visual areas generally converging more slowly and preferring narrower image distributions, relative to higher-level brain areas. Brain-optimized inference thus offers a succinct and novel method for improving reconstructions and exploring the diversity of representations across visual brain areas. Speaker: Reese Kneeland is a Ph.D. student at the University of Minnesota working in the Naselaris lab. Paper link: https://arxiv.org/abs/2312.07705

Prefrontal mechanisms involved in learning distractor-resistant working memory in a dual task

Working memory (WM) is a cognitive function that allows the short-term maintenance and manipulation of information when no longer accessible to the senses. It relies on temporarily storing stimulus features in the activity of neuronal populations. To preserve these dynamics from distraction it has been proposed that pre and post-distraction population activity decomposes into orthogonal subspaces. If orthogonalization is necessary to avoid WM distraction, it should emerge as performance in the task improves. We sought evidence of WM orthogonalization learning and the underlying mechanisms by analyzing calcium imaging data from the prelimbic (PrL) and anterior cingulate (ACC) cortices of mice as they learned to perform an olfactory dual task. The dual task combines an outer Delayed Paired-Association task (DPA) with an inner Go-NoGo task. We examined how neuronal activity reflected the process of protecting the DPA sample information against Go/NoGo distractors. As mice learned the task, we measured the overlap between the neural activity onto the low-dimensional subspaces that encode sample or distractor odors. Early in the training, pre-distraction activity overlapped with both sample and distractor subspaces. Later in the training, pre-distraction activity was strictly confined to the sample subspace, resulting in a more robust sample code. To gain mechanistic insight into how these low-dimensional WM representations evolve with learning we built a recurrent spiking network model of excitatory and inhibitory neurons with low-rank connections. The model links learning to (1) the orthogonalization of sample and distractor WM subspaces and (2) the orthogonalization of each subspace with irrelevant inputs. We validated (1) by measuring the angular distance between the sample and distractor subspaces through learning in the data. Prediction (2) was validated in PrL through the photoinhibition of ACC to PrL inputs, which induced early-training neural dynamics in well-trained animals. In the model, learning drives the network from a double-well attractor toward a more continuous ring attractor regime. We tested signatures for this dynamical evolution in the experimental data by estimating the energy landscape of the dynamics on a one-dimensional ring. In sum, our study defines network dynamics underlying the process of learning to shield WM representations from distracting tasks.

Generating parallel representations of position and identity in the olfactory system

Learning representations of specifics and generalities over time

The Insights and Outcomes of the Wellcome-funded Waiting Times Project

Waiting is one of healthcare’s core experiences. It is there in the time it takes to access services; through the days, weeks, months or years needed for diagnoses; in the time that treatment takes; and in the elongated time-frames of recovery, relapse, remission and dying.Funded by the Wellcome Trust, our project opens up what it means to wait in and for healthcare by examining lived experiences, representations and histories of delayed and impeded time.In an era in which time is lived at increasingly different and complex tempos, Waiting Times looks to understand both the difficulties and vital significance of waiting for practices of care, offering a fundamental re-conceptualisation of the relation between time and care in contemporary thinking about health, illness, and wellbeing.

Movement planning as a window into hierarchical motor control

The ability to organise one's body for action without having to think about it is taken for granted, whether it is handwriting, typing on a smartphone or computer keyboard, tying a shoelace or playing the piano. When compromised, e.g. in stroke, neurodegenerative and developmental disorders, the individuals’ study, work and day-to-day living are impacted with high societal costs. Until recently, indirect methods such as invasive recordings in animal models, computer simulations, and behavioural markers during sequence execution have been used to study covert motor sequence planning in humans. In this talk, I will demonstrate how multivariate pattern analyses of non-invasive neurophysiological recordings (MEG/EEG), fMRI, and muscular recordings, combined with a new behavioural paradigm, can help us investigate the structure and dynamics of motor sequence control before and after movement execution. Across paradigms, participants learned to retrieve and produce sequences of finger presses from long-term memory. Our findings suggest that sequence planning involves parallel pre-ordering of serial elements of the upcoming sequence, rather than a preparation of a serial trajectory of activation states. Additionally, we observed that the human neocortex automatically reorganizes the order and timing of well-trained movement sequences retrieved from memory into lower and higher-level representations on a trial-by-trial basis. This echoes behavioural transfer across task contexts and flexibility in the final hundreds of milliseconds before movement execution. These findings strongly support a hierarchical and dynamic model of skilled sequence control across the peri-movement phase, which may have implications for clinical interventions.

Learning to Express Reward Prediction Error-like Dopaminergic Activity Requires Plastic Representations of Time

The dominant theoretical framework to account for reinforcement learning in the brain is temporal difference (TD) reinforcement learning. The TD framework predicts that some neuronal elements should represent the reward prediction error (RPE), which means they signal the difference between the expected future rewards and the actual rewards. The prominence of the TD theory arises from the observation that firing properties of dopaminergic neurons in the ventral tegmental area appear similar to those of RPE model-neurons in TD learning. Previous implementations of TD learning assume a fixed temporal basis for each stimulus that might eventually predict a reward. Here we show that such a fixed temporal basis is implausible and that certain predictions of TD learning are inconsistent with experiments. We propose instead an alternative theoretical framework, coined FLEX (Flexibly Learned Errors in Expected Reward). In FLEX, feature specific representations of time are learned, allowing for neural representations of stimuli to adjust their timing and relation to rewards in an online manner. In FLEX dopamine acts as an instructive signal which helps build temporal models of the environment. FLEX is a general theoretical framework that has many possible biophysical implementations. In order to show that FLEX is a feasible approach, we present a specific biophysically plausible model which implements the principles of FLEX. We show that this implementation can account for various reinforcement learning paradigms, and that its results and predictions are consistent with a preponderance of both existing and reanalyzed experimental data.

Distinct contributions of different anterior frontal regions to rule-guided decision-making in primates: complementary evidence from lesions, electrophysiology, and neurostimulation

Different prefrontal areas contribute in distinctly different ways to rule-guided behaviour in the context of a Wisconsin Card Sorting Test (WCST) analog for macaques. For example, causal evidence from circumscribed lesions in NHPs reveals that dorsolateral prefrontal cortex (dlPFC) is necessary to maintain a reinforced abstract rule in working memory, orbitofrontal cortex (OFC) is needed to rapidly update representations of rule value, and the anterior cingulate cortex (ACC) plays a key role in cognitive control and integrating information for correct and incorrect trials over recent outcomes. Moreover, recent lesion studies of frontopolar cortex (FPC) suggest it contributes to representing the relative value of unchosen alternatives, including rules. Yet we do not understand how these functional specializations relate to intrinsic neuronal activities nor the extent to which these neuronal activities differ between different prefrontal regions. After reviewing the aforementioned causal evidence I will present our new data from studies using multi-area multi-electrode recording techniques in NHPs to simultaneously record from four different prefrontal regions implicated in rule-guided behaviour. Multi-electrode micro-arrays (‘Utah arrays’) were chronically implanted in dlPFC, vlPFC, OFC, and FPC of two macaques, allowing us to simultaneously record single and multiunit activity, and local field potential (LFP), from all regions while the monkey performs the WCST analog. Rule-related neuronal activity was widespread in all areas recorded but it differed in degree and in timing between different areas. I will also present preliminary results from decoding analyses applied to rule-related neuronal activities both from individual clusters and also from population measures. These results confirm and help quantify dynamic task-related activities that differ between prefrontal regions. We also found task-related modulation of LFPs within beta and gamma bands in FPC. By combining this correlational recording methods with trial-specific causal interventions (electrical microstimulation) to FPC we could significantly enhance and impair animals performance in distinct task epochs in functionally relevant ways, further consistent with an emerging picture of regional functional specialization within a distributed framework of interacting and interconnected cortical regions.

Learning through the eyes and ears of a child

Young children have sophisticated representations of their visual and linguistic environment. Where do these representations come from? How much knowledge arises through generic learning mechanisms applied to sensory data, and how much requires more substantive (possibly innate) inductive biases? We examine these questions by training neural networks solely on longitudinal data collected from a single child (Sullivan et al., 2020), consisting of egocentric video and audio streams. Our principal findings are as follows: 1) Based on visual only training, neural networks can acquire high-level visual features that are broadly useful across categorization and segmentation tasks. 2) Based on language only training, networks can acquire meaningful clusters of words and sentence-level syntactic sensitivity. 3) Based on paired visual and language training, networks can acquire word-referent mappings from tens of noisy examples and align their multi-modal conceptual systems. Taken together, our results show how sophisticated visual and linguistic representations can arise through data-driven learning applied to one child’s first-person experience.

The Neural Race Reduction: Dynamics of nonlinear representation learning in deep architectures

What is the relationship between task, network architecture, and population activity in nonlinear deep networks? I will describe the Gated Deep Linear Network framework, which schematizes how pathways of information flow impact learning dynamics within an architecture. Because of the gating, these networks can compute nonlinear functions of their input. We derive an exact reduction and, for certain cases, exact solutions to the dynamics of learning. The reduction takes the form of a neural race with an implicit bias towards shared representations, which then govern the model’s ability to systematically generalize, multi-task, and transfer. We show how appropriate network architectures can help factorize and abstract knowledge. Together, these results begin to shed light on the links between architecture, learning dynamics and network performance.

Relations and Predictions in Brains and Machines

Humans and animals learn and plan with flexibility and efficiency well beyond that of modern Machine Learning methods. This is hypothesized to owe in part to the ability of animals to build structured representations of their environments, and modulate these representations to rapidly adapt to new settings. In the first part of this talk, I will discuss theoretical work describing how learned representations in hippocampus enable rapid adaptation to new goals by learning predictive representations, while entorhinal cortex compresses these predictive representations with spectral methods that support smooth generalization among related states. I will also cover recent work extending this account, in which we show how the predictive model can be adapted to the probabilistic setting to describe a broader array of generalization results in humans and animals, and how entorhinal representations can be modulated to support sample generation optimized for different behavioral states. In the second part of the talk, I will overview some of the ways in which we have combined many of the same mathematical concepts with state-of-the-art deep learning methods to improve efficiency and performance in machine learning applications like physical simulation, relational reasoning, and design.

Developmentally structured coactivity in the hippocampal trisynaptic loop

The hippocampus is a key player in learning and memory. Research into this brain structure has long emphasized its plasticity and flexibility, though recent reports have come to appreciate its remarkably stable firing patterns. How novel information incorporates itself into networks that maintain their ongoing dynamics remains an open question, largely due to a lack of experimental access points into network stability. Development may provide one such access point. To explore this hypothesis, we birthdated CA1 pyramidal neurons using in-utero electroporation and examined their functional features in freely moving, adult mice. We show that CA1 pyramidal neurons of the same embryonic birthdate exhibit prominent cofiring across different brain states, including behavior in the form of overlapping place fields. Spatial representations remapped across different environments in a manner that preserves the biased correlation patterns between same birthdate neurons. These features of CA1 activity could partially be explained by structured connectivity between pyramidal cells and local interneurons. These observations suggest the existence of developmentally installed circuit motifs that impose powerful constraints on the statistics of hippocampal output.

Are place cells just memory cells? Probably yes

Neurons in the rodent hippocampus appear to encode the position of the animal in physical space during movement. Individual ``place cells'' fire in restricted sub-regions of an environment, a feature often taken as evidence that the hippocampus encodes a map of space that subserves navigation. But these same neurons exhibit complex responses to many other variables that defy explanation by position alone, and the hippocampus is known to be more broadly critical for memory formation. Here we elaborate and test a theory of hippocampal coding which produces place cells as a general consequence of efficient memory coding. We constructed neural networks that actively exploit the correlations between memories in order to learn compressed representations of experience. Place cells readily emerged in the trained model, due to the correlations in sensory input between experiences at nearby locations. Notably, these properties were highly sensitive to the compressibility of the sensory environment, with place field size and population coding level in dynamic opposition to optimally encode the correlations between experiences. The effects of learning were also strongly biphasic: nearby locations are represented more similarly following training, while locations with intermediate similarity become increasingly decorrelated, both distance-dependent effects that scaled with the compressibility of the input features. Using virtual reality and 2-photon functional calcium imaging in head-fixed mice, we recorded the simultaneous activity of thousands of hippocampal neurons during virtual exploration to test these predictions. Varying the compressibility of sensory information in the environment produced systematic changes in place cell properties that reflected the changing input statistics, consistent with the theory. We similarly identified representational plasticity during learning, which produced a distance-dependent exchange between compression and pattern separation. These results motivate a more domain-general interpretation of hippocampal computation, one that is naturally compatible with earlier theories on the circuit's importance for episodic memory formation. Work done in collaboration with James Priestley, Lorenzo Posani, Marcus Benna, Attila Losonczy.

Investigating semantics above and beyond language: a clinical and cognitive neuroscience approach

The ability to build, store, and manipulate semantic representations lies at the core of all our (inter)actions. Combining evidence from cognitive neuroimaging and experimental neuropsychology, I study the neurocognitive correlates of semantic knowledge in relation to other cognitive functions, chiefly language. In this talk, I will start by reviewing neuroimaging findings supporting the idea that semantic representations are encoded in distributed yet specialized cortical areas (1), and rapidly recovered (2) according to the requirement of the task at hand (3). I will then focus on studies conducted in neurodegenerative patients, offering a unique window on the key role played by a structurally and functionally heterogeneous piece of cortex: the anterior temporal lobe (4,5). I will present pathological, neuroimaging, cognitive, and behavioral data illustrating how damages to language-related networks can affect or spare semantic knowledge as well as possible paths to functional compensation (6,7). Time permitting, we will discuss the neurocognitive dissociation between nouns and verbs (8) and how verb production is differentially impacted by specific language impairments (9).

A specialized role for entorhinal attractor dynamics in combining path integration and landmarks during navigation

During navigation, animals estimate their position using path integration and landmarks. In a series of two studies, we used virtual reality and electrophysiology to dissect how these inputs combine to generate the brain’s spatial representations. In the first study (Campbell et al., 2018), we focused on the medial entorhinal cortex (MEC) and its set of navigationally-relevant cell types, including grid cells, border cells, and speed cells. We discovered that attractor dynamics could explain an array of initially puzzling MEC responses to virtual reality manipulations. This theoretical framework successfully predicted both MEC grid cell responses to additional virtual reality manipulations, as well as mouse behavior in a virtual path integration task. In the second study (Campbell*, Attinger* et al., 2021), we asked whether these principles generalize to other navigationally-relevant brain regions. We used Neuropixels probes to record thousands of neurons from MEC, primary visual cortex (V1), and retrosplenial cortex (RSC). In contrast to the prevailing view that “everything is everywhere all at once,” we identified a unique population of MEC neurons, overlapping with grid cells, that became active with striking spatial periodicity while head-fixed mice ran on a treadmill in darkness. These neurons exhibited unique cue-integration properties compared to other MEC, V1, or RSC neurons: they remapped more readily in response to conflicts between path integration and landmarks; they coded position prospectively as opposed to retrospectively; they upweighted path integration relative to landmarks in conditions of low visual contrast; and as a population, they exhibited a lower-dimensional activity structure. Based on these results, our current view is that MEC attractor dynamics play a privileged role in resolving conflicts between path integration and landmarks during navigation. Future work should include carefully designed causal manipulations to rigorously test this idea, and expand the theoretical framework to incorporate notions of uncertainty and optimality.

Human see, human do? Tool use representations during picture viewing, pantomiming and real grasping

Predictive modeling, cortical hierarchy, and their computational implications

Predictive modeling and dimensionality reduction of functional neuroimaging data have provided rich information about the representations and functional architectures of the human brain. While these approaches have been effective in many cases, we will discuss how neglecting the internal dynamics of the brain (e.g., spontaneous activity, global dynamics, effective connectivity) and its underlying computational principles may hinder our progress in understanding and modeling brain functions. By reexamining evidence from our previous and ongoing work, we will propose new hypotheses and directions for research that consider both internal dynamics and the computational principles that may govern brain processes.

Geometry of concept learning

Understanding Human ability to learn novel concepts from just a few sensory experiences is a fundamental problem in cognitive neuroscience. I will describe a recent work with Ben Sorcher and Surya Ganguli (PNAS, October 2022) in which we propose a simple, biologically plausible, and mathematically tractable neural mechanism for few-shot learning of naturalistic concepts. We posit that the concepts that can be learned from few examples are defined by tightly circumscribed manifolds in the neural firing-rate space of higher-order sensory areas. Discrimination between novel concepts is performed by downstream neurons implementing ‘prototype’ decision rule, in which a test example is classified according to the nearest prototype constructed from the few training examples. We show that prototype few-shot learning achieves high few-shot learning accuracy on natural visual concepts using both macaque inferotemporal cortex representations and deep neural network (DNN) models of these representations. We develop a mathematical theory that links few-shot learning to the geometric properties of the neural concept manifolds and demonstrate its agreement with our numerical simulations across different DNNs as well as different layers. Intriguingly, we observe striking mismatches between the geometry of manifolds in intermediate stages of the primate visual pathway and in trained DNNs. Finally, we show that linguistic descriptors of visual concepts can be used to discriminate images belonging to novel concepts, without any prior visual experience of these concepts (a task known as ‘zero-shot’ learning), indicated a remarkable alignment of manifold representations of concepts in visual and language modalities. I will discuss ongoing effort to extend this work to other high level cognitive tasks.

The power of structured representations (and how to learn them)

Convex neural codes in recurrent networks and sensory systems

Neural activity in many sensory systems is organized on low-dimensional manifolds by means of convex receptive fields. Neural codes in these areas are constrained by this organization, as not every neural code is compatible with convex receptive fields. The same codes are also constrained by the structure of the underlying neural network. In my talk I will attempt to provide answers to the following natural questions: (i) How do recurrent circuits generate codes that are compatible with the convexity of receptive fields? (ii) How can we utilize the constraints imposed by the convex receptive field to understand the underlying stimulus space. To answer question (i), we describe the combinatorics of the steady states and fixed points of recurrent networks that satisfy the Dale’s law. It turns out the combinatorics of the fixed points are completely determined by two distinct conditions: (a) the connectivity graph of the network and (b) a spectral condition on the synaptic matrix. We give a characterization of exactly which features of connectivity determine the combinatorics of the fixed points. We also find that a generic recurrent network that satisfies Dale's law outputs convex combinatorial codes. To address question (ii), I will describe methods based on ideas from topology and geometry that take advantage of the convex receptive field properties to infer the dimension of (non-linear) neural representations. I will illustrate the first method by inferring basic features of the neural representations in the mouse olfactory bulb.

Neural Dynamics of Cognitive Control

Cognitive control guides behavior by controlling what, where, and how information is represented in the brain. Perhaps the most well-studied form of cognitive control has been ‘attention’, which controls how external sensory stimuli are represented in the brain. In contrast, the neural mechanisms controlling the selection of representations held ‘in mind’, in working memory, are unknown. In this talk, I will present evidence that the prefrontal cortex controls working memory by selectively enhancing and transforming the contents of working memory. In particular, I will show how the neural representation of the content of working memory changes over time, moving between different ‘subspaces’ of the neural population. These dynamics may play a critical role in controlling what and how neural representations are acted upon.