Affiliation: New York University

Bio: Dora Angelaki is a Professor at the Center for Neural Science and the Tandon school of Engineering at New York University. She holds a diploma and Ph.D. degrees in electrical and biomedical engineering from the National Technical University of Athens, Greece, and the University of Minnesota. interested in understanding the principles that make our brain so much better than man-made machines and AI. She uses naturalistic foraging tasks that combine uncertainty, spatial navigation, decision-making and episodic memory to understand inference in the brain. She explores how task-relevant latent variables and multisensory signals flow dynamically across brain areas to generate perception and cognition, how hierarchical causal inference is implemented in the brain, how beliefs propagate through the network, and how internal states modulate this information flow.

Learn more

Affiliation: Biozentrum, University of Basel, and Friedrich Miescher Institute, Basel, Switzerland

Title: Professor of Computational Neuroscience

Bio: Silvia Arber is a Swiss neuroscientist holding a Full Professor position at the Biozentrum of the University of Basel and a Senior Group leader appointment at the Friedrich Miescher Institute for Biomedical Research, in Basel, Switzerland. She is recognized for her work on the organization and function of neuronal circuits controlling movement. Arber received her PhD from the University of Basel (1996) and carried out her postdoctoral work at Columbia University in New York (1996-2000). She received numerous prizes and honors, including most recently the Brain Prize from the Lundbeck Foundation (2022), and being elected international member of the National Academy of Sciences (USA).

Talk Title: Licensing actions through basal ganglia and brainstem circuits

Abstract: Movement is the behavioral output of the nervous system. Recent work has begun to dissect the cell types and circuits involved in the regulation of diverse forms of body movement. The brainstem is an important switchboard between motor system planning centers and the circuit elements engaged in movement execution in the spinal cord. This talk will highlight circuit organization and neuronal dynamics at the intersection between basal ganglia and action output centers for body movements in the brainstem. Using circuit tracing approaches, mouse genetics and a variety of technologies to record from identified neurons in combination with tracking behavior, we have identified brainstem centers dedicated to the generation of full body movements including locomotion and posture, as well as regions involved in the construction of complex forelimb movements. Strikingly, dedicated circuit modules defined by cell types are at the core of generating these diverse forms of movement with high specificity. The regulation of these specific cell types by highly specific synaptic inputs from higher motor centers including cortex and basal ganglia as well as intra-brainstem processing of information are important ingredients in the process of generating behavioral specificity as well as flexibility.

Learn more

Affiliation: Oxford University

Title: Professor of Computational Neuroscience

Bio: Tim Behrens works at Oxford and UCL, where he thinks about how the brain organises knowledge for flexible behaviour.

Talk Title: Representing the structure of sequences

Abstract: I am mostly going to talk about 1 study about how a behavioural schema is represented in frontal cortex.   If there is time, I might also show how this is different to how a behavioural schema is represented in entorhinal cortex.

Learn more

Keynote Speaker

Bio: Sebastien Bubeck is a Partner Research Manager at Microsoft Research (MSR). He won multiple awards for his machine learning results around robustness and optimization. Most recently, he has been interested in understanding how intelligence emerges in large language models. With collaborators at MSR, he wrote the Sparks of AGI paper studying the emergent abilities of GPT-4. This work was covered in The New York Times, Wired, This American Life, and more.

Bio: Beth Buffalo is a neuroscientist whose lab focuses on understanding the neural mechanisms that support learning and memory. Dr. Buffalo received a B.A. in Philosophy from Wellesley College and a Ph.D. in Neurosciences from the University of California, San Diego. Dr. Buffalo joined the faculty at Emory University in 2005, and then moved her lab to the University of Washington in 2013, where she currently serves as the Wayne E. Crill Endowed Professor and Chair of Physiology and Biophysics. She is also a Core Faculty member of the Washington National Primate Research Center. Dr. Buffalo has received several awards for her research including the Troland Research Award from the National Academy of Sciences for her innovative, multidisciplinary study of the hippocampus and the neural basis of memory. She is an elected member of the National Academy of Sciences.

Talk Title: Saccadic Eye Movements Structure Activity in the Primate Hippocampus

Affiliation: Champalimaud Neuroscience Programme; Champalimaud Foundation

Title: Principal Investigator, Laboratory of Sensorimotor Integration

Bio: Eugenia Chiappe received a licenciatura in Biological Sciences from the University of Buenos Aires (Argentina) before pursuing her PhD as an HHMI predoctoral fellow at The Rockefeller University. As a postdoctoral fellow at HHMI’s Janelia Research Campus, she played a pivotal role in establishing simultaneous recordings of neural activity and behavior in Drosophila. Her lab at the Champalimaud Neuroscience Programme (Portugal) focuses on investigating how the brain integrates motor and sensory information for self-motion estimation and movement control. The Chiappe lab’s research has revealed the integration of multimodal information in visual pathways for self-motion estimation (Fujiwara et al., Nature Neuroscience, 2017) and the flexible recruitment mechanisms of these pathways for behavior control (Fujiwara et al., Neuron, 2022). Furthermore, using virtual reality tools, the Chiappe lab has demonstrated how visual feedback controls exploratory walking (Cruz et al., Current Biology, 2021). Eugenia has received numerous honors, including multiple funding grants from ERC.

Talk Title: Recurrent circuit interactions for invariant self-motion estimation in walking Drosophila

Abstract: Movement control depends on continuously mapping sensory activity to movement parameters while comparing postural dynamics with animal goals. The brain’s capacity to estimate self-motion invariant to the state of environment and the body is pivotal to this mapping. Yet, little is known about the configuration of circuits supporting an invariant self-motion estimate, and how this information is used to control locomotion in the proper context.
Using quantitative tools to analyze behavior, physiology, and anatomy, we have begun to uncover the structure and function of optic flow-sensitive pathways for self-rotation estimation and steering control. We discovered layered pathways with divergent outputs leading to executive centers or deep premotor regions of the fly brain. A characteristic feature of the layered network is competitive lateral disinhibition, playing a critical role on detecting rotational optic flow regardless of the fly’s forward speed. In addition, the network is characterized by strong reciprocal interactions, some of which contribute to multimodal integration for self-rotation estimation invariant to the visual structure of the environment. Finally, we found that a continuous bidirectional interaction circuits within the brain and ventral nerve cord (VNC, the insect analogue of the spinal cord), operating at multiple time scales, facilitates rapid, steering control specifically when the fly walks at high speed.
Our findings reveal the configuration of central networks processing optic flow information robustly with finely tuned inhibitory mechanisms, enabling premotor networks to orchestrate the interplay between mechanical stability and flexible course control via an accurate estimate of body rotations. This orchestration, observed in natural behaviors like exploration, adapts dynamically to specific motor contexts.

Affiliation: New York University & Flatiron Institute

Bio: SueYeon Chung is an Assistant Professor in the Center for Neural Science at NYU, with a joint appointment in the Center for Computational Neuroscience at the Flatiron Institute. She is also an affiliated faculty member at the Center for Data Science and Cognition and Perception Program at NYU. Prior to joining NYU, she was a Postdoctoral Fellow in the Center for Theoretical Neuroscience at Columbia University, and BCS Fellow in Computation at MIT. Before that, she received a Ph.D. in applied physics at Harvard University, and a B.A. in mathematics and physics at Cornell University.

Talk Title: Multi-level theory of neural representations:
Capacity of neural manifolds in biological and artificial neural networks

Abstract: A central goal in neuroscience is to understand how orchestrated computations in the brain arise from the properties of single neurons and networks of such neurons. Answering this question requires theoretical advances that shine a light on the ‘black box’ of representations in neural circuits. In this talk, we will demonstrate theoretical approaches that help describe how cognitive task implementations emerge from the structure in neural populations and from biologically plausible neural networks.
We will introduce a new statistical mechanical theory that connects geometric structures that arise from neural responses (i.e., neural manifolds) to the neural representation’s efficiency in implementing a task. In particular, this theory describes how many neural manifolds can be represented (or ‘packed’) in the neural activity space while they can be linearly decoded by a downstream readout neuron. The intuition from this theory is remarkably simple: like a sphere packing problem in physical space, we can encode many “neural manifolds” into the neural activity space if these manifolds are small and low-dimensional, and vice versa.
Next, we will describe how such an approach can, in fact, open the ‘black box’ of distributed neuronal circuits in a range of settings, such as experimental neural datasets and artificial neural networks. In particular, our method overcomes the limitations of traditional dimensionality reduction techniques, as it operates directly on the high-dimensional representations rather than relying on low-dimensionality assumptions for visualization. Furthermore, this method allows for simultaneous multi-level analysis, by measuring geometric properties in neural population data and estimating the amount of task information embedded in the same population. This geometric approach is general and can be used across different brain areas and task modalities, as demonstrated in our works and others.
Finally, we will discuss our recent efforts to fully extend this multi-level description of neural populations by (1) investigating how single neuron properties shape the representation geometry in early sensory areas and by (2) understanding how task-implementing neural manifolds emerge in downstream areas in biological and artificial networks. By extending our mathematical toolkit for analyzing representations underlying complex neuronal networks, we hope to contribute to the long-term challenge of understanding the neuronal basis of tasks and behaviors.

Affiliation: University of California Los Angeles

Talk Title: Engaged decision-makers align spontaneous movements to stereotyped task demands.

Abstract: Neural activity during sensory-guided decision-making tasks is strongly modulated by the animal’s movements. Although the impact of movements on neural activity is now well-documented in multiple species, the relationship between these movements and behavioral performance remains unclear. To understand this relationship, we first tested whether the magnitude of animal movements (assessed by the motion energy of 28 DeepLabCut-labeled body parts) was correlated with its performance on a perceptual decision-making task. No strong relationship was present, suggesting that task performance is not affected by the magnitude of movements. We then tested if performance instead depends on  movement timing and trajectory. We therefore partitioned the movements into two groups: task-aligned movements that were well predicted by task events (such as the onset of the sensory stimulus or the animal’s choice) and task independent movements (TIM) that occurred independently of task events. TIM had a reliable, inverse correlation with performance in 8/9 head-fixed mice and 4/4 freely moving rats. This argues that certain movements, defined by their timing and trajectories relative to task events, might indicate periods of engagement or disengagement in the task. To confirm this, we compared TIM to the latent behavioral states recovered by a hidden Markov model with Bernoulli generalized linear model observations (GLM-HMM) and found these, again, to be inversely correlated. Finally, we examined the impact of these behavioral states on neural activity measured with wide field calcium imaging. Our linear encoding model could account for more overall variance in neural activity in the disengaged state, likely because of the preponderance of TIM during that time. Interestingly, the engaged state was associated with widespread increased activity, particularly during the delay period. Taken together, these findings suggest that TIM is informative about an animal’s internal state of engagement, and that the movements and state together have a major impact on neural activity, especially during the most cognitively demanding moments in a decision. 

Affiliation: Imperial College London

Talk Title: De novo motor learning creates structure in neural activity space that shapes adaptation

Abstract: Animals can quickly adapt learned movements in response to external perturbations. Motor adaptation is likely influenced by an animal’s existing movement repertoire, but the nature of this influence is unclear. Long-term learning causes lasting changes in neural connectivity which determine the activity patterns that can be produced. Here, we sought to understand how a neural population’s activity repertoire, acquired through long-term learning, affects short-term adaptation by modeling motor cortical neural population dynamics during de novo learning and subsequent adaptation using recurrent neural networks. We trained these networks on different motor repertoires comprising varying numbers of movements. Networks with multiple movements had more constrained and robust dynamics, which were associated with more defined neural ‘structure’-organization created by the neural population activity patterns corresponding to each movement. This structure facilitated adaptation, but only when small changes in motor output were required, and when the structure of the network inputs, the neural activity space, and the perturbation were congruent. These results highlight trade-offs in skill acquisition and demonstrate how prior experience and external cues during learning can shape the geometrical properties of neural population activity as well as subsequent adaptation.

Affiliation: Allen Institute for Neural Dynamics

Talk Title: Two types of locus coeruleus norepinephrine neurons drive reinforcement learning

Affiliation: Harvard University

Bio: Sandeep Robert Datta is a Professor in the Department of Neurobiology at Harvard Medical School. He obtained a Bachelor of Science degree in Molecular Biochemistry and Biophysics from Yale University, and obtained an M.D./Ph.D degree from Harvard University. After working as a postdoctoral fellow at Columbia University with Richard Axel, he joined HMS Neurobiology where his lab focuses on understanding how sensory cues — particularly odors — are detected by the nervous system, how the brain transforms information about the presence of salient sensory cues into patterns of motivated action, and how the brain uses behavior to usefully interact with the world. Dr. Datta is an Associate Member of the Broad Institute, and is a Principal Investigator in the Italian Institute of Technology/Harvard Medical School joint program in the neurosciences. 

Affiliation: Stanford University

Bio: I’m a postdoc with Krishna V. Shenoy and David Sussillo in the Neural Prosthetic Systems Laboratory (NPSL) at Stanford University. I’m interested in neural population dynamics in the context of learning and memory in cortical networks. How do ensembles of neurons balance flexibility and robustness for new and previously learned computations? I will be starting a Senior Scientist position at the Allen Institute for Neural Dynamics in January 2024.

Talk Title: Flexible multitask computation in recurrent networks utilizes shared dynamical motifs

Abstract: Flexible computation is a hallmark of intelligent behavior. Yet, little is known about how neural networks contextually reconfigure for different computations. Humans are able to perform a new task without extensive training, presumably through the composition of elementary processes that were previously learned. Cognitive scientists have long hypothesized the possibility of a compositional neural code, where complex neural computations are made up of constituent components; however, the neural substrate underlying this structure remains elusive in biological and artificial neural networks. Here we identified an algorithmic neural substrate for compositional computation through the study of multitasking artificial recurrent neural networks. Dynamical systems analyses of networks revealed learned computational strategies that mirrored the modular subtask structure of the task-set used for training. Dynamical motifs such as attractors, decision boundaries and rotations were reused across different task computations. For example, tasks that required memory of a continuous circular variable repurposed the same ring attractor. We show that dynamical motifs are implemented by clusters of units and are reused across different contexts, allowing for flexibility and generalization of previously learned computation. Lesioning these clusters resulted in modular effects on network performance: a lesion that destroyed one dynamical motif only minimally perturbed the structure of other dynamical motifs. Finally, modular dynamical motifs could be reconfigured for fast transfer learning. After slow initial learning of dynamical motifs, a subsequent faster stage of learning reconfigured motifs to perform novel tasks. This work contributes to a more fundamental understanding of compositional computation underlying flexible general intelligence in neural systems. We present a conceptual framework that establishes dynamical motifs as a fundamental unit of computation, intermediate between the neuron and the network. As more whole brain imaging studies record neural activity from multiple specialized systems simultaneously, the framework of dynamical motifs will guide questions about specialization and generalization across brain regions.

Affiliation: Princeton University

Bio: Tatiana Engel is an Assistant Professor at the Princeton Neuroscience Institute. She uses computational and theoretical approaches to investigate how coordinated activity arises from distributed neural circuitry to drive behavioral and cognitive functions. Her lab develops theory, models, and data analysis methods that leverage newly available large-scale activity recordings from behaving animals to uncover mechanisms of brain function. She also participates in a large-scale collaboration of experimental and theoretical neuroscientists, the International Brain Laboratory.

Talk Title: Latent circuit inference from heterogeneous neural responses during cognitive tasks

Abstract: Higher cortical areas carry a wide range of signals supporting complex goal-directed behavior, which are mixed in heterogeneous responses of single neurons tuned to multiple task variables. Current statistical methods can identify representations of task variables in neural population activity, but do not provide causal mechanisms to explain how low-dimensional activity drives behavior. We develop the latent circuit model, a dimensionality reduction approach in which task variables interact via low-dimensional recurrent connectivity to produce behavioral output. Applied to recurrent neural networks trained to perform a context-dependent decision-making task, the latent circuit model uncovered a suppression mechanism in which contextual representations inhibit irrelevant sensory responses. We validate this mechanism by confirming the behavioral effects of patterned connectivity perturbations predicted by the latent circuit model. We find similar suppression of irrelevant sensory responses in the prefrontal cortex of monkeys performing the same task. Our results show that incorporating causal interactions among task variables is critical for identifying behaviorally relevant computations. 

Affiliation: Massachusetts Institute of Technology  
Bio: Ila Fiete is a Professor in the Department of Brain and Cognitive Sciences and an Associate Member of the McGovern Institute at MIT. She obtained her undergraduate degrees in Physics and Mathematics at the University of Michigan and her M.A. and Ph.D. in Physics at Harvard, under the guidance of Sebastian Seung at MIT (co-advised by Daniel Fisher). Her postdoctoral work was at the Kavli Institute for Theoretical Physics at Santa Barbara, and at Caltech, where she was a Broad Fellow. She was subsequently on the faculty of the University of Texas at Austin in the Center for Learning and Memory. Ila Fiete is an HHMI Faculty Scholar. She has been a CIFAR Senior Fellow, a McKnight Scholar, an ONR Young Investigator, an Alfred P. Sloan Foundation Fellow, and a Searle Scholar.

Talk Title: A theory of hippocampal episodic memory unified with and enabled by prestructured spatial representations

Affiliation: Friedrich Miescher Institute for Biomedical Research (FMI)
Bio: Rainer Friedrich is a Senior Group Leader and Professor at the Friedrich Miescher Institute for Biomedical Research (FMI), Basel, Switzerland. Previously, he was trained in neuroscience and biophysics (University of Freiburg; Brock University; MPI for Developmental Biology; Caltech) and head of a junior research group at the MPI for Medical Research. His research links experiments to theory to understand how populations of neurons interact to perform computations underlying intelligent behavior. Previous work contributed significantly to the current picture of olfactory processing in vertebrates and to the development of zebrafish as a model organism in systems neuroscience. Recent work combines EM-based neuronal circuit reconstruction with large-scale activity measurements and behavioral experiments in adult zebrafish to obtain mechanistic insights into neuronal computations underlying memory and cognition. Friedrich is member of EMBO and received awards including the Delwart Research Prize and Minna-James-Heinemann award.
Talk Title: Computational functions of neuronal assemblies in a synaptically balanced memory network
Abstract: Memory and cognition are thought to depend on the formation of neuronal assemblies that modify population dynamics to support storage and classification of information. However, the direct anatomical detection of such assemblies remains a major challenge and their computational consequences are not well understood in biologically realistic networks. We address these questions using telencephalic area Dp of adult zebrafish, the homolog of piriform cortex, as a model. Training of zebrafish in an odor discrimination task modified both excitatory and inhibitory components of odor representations in Dp. However, contrary to predictions of classical models, we found no evidence for stable attractor states. Rather, during odor stimulation, Dp enters a state of precise synaptic balance and generates variable activity. Analyses of population activity revealed signatures of assemblies containing excitatory and inhibitory neurons (E-I assemblies). Optogenetic manipulations of different interneuron types further support this scenario and highlight the importance of specific inhibitory interactions in balanced memory netwdorks. Computational analyses indicate that information is represented on manifolds in neural state space with nonlinear geometry. EM-based circuit reconstructions are ongoing to directly analyze structured connectivity in Dp. In summary, our results show that E-I assemblies can establish manifold representations in balanced memory networks and indicate that such representations mediate memory and pattern classification in Dp/piriform cortex. Unlike stable attractor states, such representations support fast and continuous pattern classification, which is important for higher-order learning and cognition. 

Affiliation: Stanford University

Affiliation: Allen Institute for Neural Dynamics

Bio: Marina is an Assistant Investigator at the Allen Institute where she leads a team of scientists and engineers to design, curate, analyze, interpret, and share standardized neurophysiological datasets. She contributed to the generation of a large-scale dataset that allows investigation of the influence of experience, expectation, and task engagement on neural activity in excitatory and inhibitory neurons in the visual cortex (brain-map.org/explore/circuits/visual-behavior-2p), which led to the discovery that VIP inhibitory neurons are modulated by stimulus novelty. Her recent work aims to combine longitudinal 2-photon calcium imaging with spatial transcriptomics to study cell type specific circuit dynamics during learning and novelty processing.

Talk Title: Stimulus novelty uncovers functional diversity in visual cortical circuits 

Affiliation: University College London

Bio: I studied mathematics at Cambridge University, did a PhD in robotics at UCL, then moved to Rutgers University in the United States for postdoctoral work in neuroscience. Before returning to UCL in 2012, I was Associate Professor of Neuroscience at Rutgers, and Professor of Neurotechnology at Imperial College London. I am currently Professor of Quantitative Neuroscience in the UCL Institute of Neurology, and together with Matteo Carandini direct the Cortexlab.

Talk Title: Secondary motor cortex drives perseverative behavior in mice

Abstract: Perseveration — repeatedly performing one action even though other choices would lead to larger rewards – is a very common animal behavior.  Neither the purpose nor neuronal mechanisms of perseveration are understood. One hypothesis is that performing the same action repeatedly allows quick and efficient execution, because preparatory activity can be established in appropriate brain circuits. If so, there should exist neurons somewhere in the brain whose activity correlates with perseveration and rapid movement execution, and whose inhibition suppresses both.
We trained mice to perform a “two-armed bandit” task, responding to a tone by turning a wheel in one of two directions.  Reward was more probable for one turn direction, with the optimal direction switching in blocks.  Mouse behavior was best fit by learning models including both reward seeking and perseveration. Reaction times were fastest when choices were repeated. Recordings across the forebrain using Neuropixels electrodes revealed correlates of ongoing choices and rewards in many regions, but pre-tone predictions of upcoming choices only in secondary motor cortex (MOs). MOs activity built up over multiple choices to the same side, and predicted rapid reaction times. Optogenetically inhibiting MOs reduced perseveration and slowed reaction times.  In contrast, inhibiting medial prefrontal cortex slowed reward-based updating of the optimal choice side but did not affect perseveration.
We hypothesize that in this task, activity in MOs subserves motor preparation.  Mice plan their movement in advance of the tone, and MOs activity in the pre-tone period creates a neural state in which the tone quickly releases an action to the chosen side.  This hypothesis is further supported by results in a different task, where the appropriate choice cannot be determined prior to the done, and MOs does not predict upcoming choices.

Affiliation: Janelia Research Campus

Bio: Ann Hermundstad is a Group Leader in Computation & Theory at Janelia Research Campus. She studies how the brain creates and uses adaptive sensorimotor representations to generate flexible behavior. Her lab uses a combination of theory, modeling, and data analysis to explore how neural circuits can do this efficiently and flexibly, and works in close collaboration with experimentalists to test these ideas in biological systems. Ann received her PhD in physics from UC Santa Barbara. Before starting her lab at Janelia in 2016, she did a postdoc in theoretical neuroscience with Vijay Balasubramanian, initially at École Normale Supérieure in Paris and later at UPenn in Philadelphia.

Talk Title: Structural priors for rapid learning

Abstract: Many flexible behaviors are thought to rely on internal representations of an animal’s spatial relationship to its environment and of the consequences of its actions in that environment. While such representations—e.g. of head direction and value—have been extensively studied, how they are constructed and combined to guide behavior is not well understood. We use a combination of theory and modeling, together with behavioral, neural, and connectomic data, to study these questions in a visual learning paradigm that requires fruit flies to simultaneously map a new environment and infer goals within that environment. To perform this task, animals must modify their behavior within a matter of minutes, an ability that we think relies on genetically-specified circuits that incorporate structured relationships between an animal and its surroundings, and on continually-evolving internal representations carried within these circuits. I will discuss how small circuits can accurately maintain and modify these representations in the face of strong anatomical constraints, and how these constraints, in turn, can shed light on rapid learning and behavioral flexibility in biological and artificial agents more broadly. 

Affiliation: Max Planck Florida

Bio: Hidehiko Inagaki is a Research Group Leader at the Max Planck Florida Institute for Neuroscience in Jupiter, FL. He received his B.S. in Biochemistry and Biophysics from the University of Tokyo, and his Ph.D. in Biology from California Institute of Technology, where he studied the neuronal mechanisms of internal state control in Drosophila. As a Helen Hay Whitney Postdoctoral Fellow in Karel Svoboda lab at Janelia Research Campus, HHMI, he studied the neuronal mechanisms of short-term memory in mice. He started the Inagaki lab in Sep 2019 to study neuronal mechanisms of motor and cognitive functions leveraging both molecular and systems neuroscience methods. Hidehiko is a recipient of the Searle Scholar Award, the Peter and Patricia Gruber International Research Award in Neuroscience, Klingenstein-Simons Foundation Fellow Award, NIH Director’s New Innovator Award, and McKnight Scholar Award.

Talk Title: Neocortical mechanisms of motor learning

Abstract: Neocortical spiking dynamics control aspects of behavior, yet how these dynamics emerge during motor learning remains elusive. Activity-dependent synaptic plasticity is likely a key mechanism, as it reconfigures network architectures that govern neural dynamics. Here, we examined how the mouse premotor cortex acquires its well-characterized neural dynamics that control movement timing, specifically lick timing. To probe the role of synaptic plasticity, we have genetically manipulated proteins essential for major forms of synaptic plasticity, Ca2+/calmodulin-dependent protein kinase II (CaMKII) and Cofilin, in a region and cell-type-specific manner. Transient inactivation of CaMKII in the premotor cortex blocked learning of new lick timing without affecting the execution of learned action or ongoing spiking activity. Furthermore, among the major glutamatergic neurons in the premotor cortex, CaMKII and Cofilin activity in pyramidal tract (PT) neurons but not intratelencephalic (IT) neurons is necessary for learning. High-density electrophysiology in the premotor cortex uncovered that neural dynamics anticipating licks are progressively shaped during learning, which explains the change in lick timing. Such directional reconfiguration in behaviorally-relevant dynamics is impeded by CaMKII manipulation in PT neurons. Altogether, the activity of plasticity-related proteins in PT neurons is uniquely required for sculpting neocortical dynamics to learn new behavior. 

Affiliation: Massachusetts Institute of Technology

Bio: Mehrdad Jazayeri is an associate professor and director of education in MIT’s Department of Brain and Cognitive Sciences and an investigator at McGovern Institute. He received his BSc in electrical engineering majoring in telecommunications from Sharif University of Technology in Tehran, Iran. He then completed his MS in physiology at the University of Toronto, and his PhD in neuroscience at New York University.

Talk Title: Vector production via mental navigation in the entorhinal cortex

Abstract: A cognitive map is a suitably structured representation that enables an agent to perform novel computations using prior experience, for instance, planning a new route in a familiar space. Recent work in mammals has found direct evidence for such structured representations in the presence of exogenous sensory inputs in both spatial and non-spatial domains. Here, we test a foundational postulate of the original cognitive map theory that cognitive maps are recruited endogenously during mental navigation without external input. We recorded from the entorhinal cortex of monkeys in a mental navigation task that required animals to use a joystick to produce one-dimensional vectors between pairs of visual landmarks without sensory feedback about the intermediate landmarks. Animals’ ability to perform the task and generalize to new pairs indicated that they relied on a structured representation of the landmarks. Task-modulated neurons exhibited periodicity and ramping that matched the temporal structure of the landmarks. Neuron pairs with high periodicity scores had invariant cross-correlation structure, a signature of grid cell continuous attractor states. A basic continuous attractor network model of path integration augmented with a Hebbian learning mechanism provided an explanation of how the system endogenously recalls landmarks. The model also made an unexpected prediction that endogenous landmarks transiently slow down path integration, reset the dynamics, and thereby, reduce variability. Remarkably, this prediction was borne out of a reanalysis of behavior. Together, our findings connect the structured activity patterns in the entorhinal cortex to the endogenous recruitment of a cognitive map during mental navigation. 

Affiliation: Friedrich Miescher Institute for Biomedical Research (FMI)

Bio: Georg Keller is a neuroscientist at the Friedrich Miescher Institute for Biomedical Research in Basel Switzerland. His lab aims to understand the mechanisms that give rise to conscious perception and how these mechanisms are altered in psychiatric diseases.

Talk Title: Cortical circuits for predictive processing

Abstract: Predictive processing is a computational framework for understanding cortical function. In my talk, I will provide an overview of our current proposal for how predictive processing may be implemented in cortical circuits, and how emerging evidence challenges some of these ideas.   

Affiliation: Allen Institute

Bio: Stefan Mihalas joined the Allen Institute in 2011 from Johns Hopkins University, where he was a postdoctoral fellow in neuroscience and subsequently an associate research scientist. As a computational neuroscientist, Mihalas has worked on models of both molecular and systems neuroscience including nervous system development, synaptic plasticity, minimalistic spiking neuron models, self-organized criticality, visual attention and figure ground segregation. His current research interests are aimed at building models to elucidate how large networks of interacting neurons produce cognitive behaviors. At the Allen Institute, Mihalas integrates anatomical and physiological connectivity data to generate models of visual perception in the mouse. To this end, he works to build a series of models of increasing complexity for both individual components, i.e., neurons, synapses, and microcircuits, as well as for large portions of the entire system. This series of models will be compared to the simplified theoretical predictions from statistical physics, information theory and computer vision. Mihalas received his Diploma in physics and M.S. in mathematics from West University of Timisoara in Romania. He received his Ph.D. in physics from the California Institute of Technology.

Talk Title: Computing with dynamic synapses

Abstract: How biological neural networks implement computations is shaped by the tasks they need to solve and constrained by the dynamics of the components used. Yet often models which are trained to solve tasks are based on artificial neural networks, which have very simple components. One such oversimplified component is synapses, which in recurrent neural networks (RNNs) have their weights frozen after training, relying on an internal state stored in neuronal activity to hold task-relevant information. In biological networks, in addition to long-timescale rewiring, synapses are subject to significant modulation at faster timescales. In this presentation I will show a few computational properties of networks with dynamic synapses, and compare activity and behavior of networks with dynamic synapses to the observations of neural activity and behavior of mice trained on a detection of change task.

Specifically, I am going to show that networks with dynamic synapses:

1. Have fundamentally different dynamics than RNNs
2. Can solve a wide battery of tasks with much fewer trained parameters than RNNs
3. When trained on a detection of change task, they automatically fit neuronal activity and mouse behavior better than RNNs
4. Can reproduce all the novelty responses observed in the detection of change task in mice
5. Have functional cell types naturally emerge from learning.

Given these observations, I will conclude that recurrent neural networks have significant limitations as models of computation in biological networks, especially for tasks with long time scales.

Affiliation: Norwegian University of Science and Technology

Bio: Edvard Moser studies neural network codes for space, time and memory. His work, conducted in collaboration with May-Britt Moser, includes the discovery of grid cells in the brain´s position system. Their current focus is on unravelling how space and time emerge in collective activity of large neuronal populations, an endeavour significantly boosted by Neuropixels probes and 2-photon miniscopes for freely-moving rodents.
Moser received his initial training with Per Andersen at the University of Oslo, Richard Morris at the University of Edinburgh, and John O’Keefe at University College London. In 1996 he accepted a professorship at NTNU. With May-Britt Moser he founded the Centre for the Biology of Memory in 2002, the Kavli Institute for Systems Neuroscience in 2007, the Centre for Neural Computation in 2013, and the Centre for Algorithms in the Cortex in 2023. The Mosers have received numerous awards, including the 2014 Nobel Prize in Medicine or Physiology.

Talk Title: Computing space in networks of the entorhinal cortex

Abstract: Space is mapped by complex neural networks in the hippocampus and medial entorhinal cortex. These brain areas contain specialized position-coding cell types, including place cells in the hippocampus and grid cells in the medial entorhinal cortex. I will show how new technologies such as Neuropixels probes allow the dynamics of thousands of place and grid cells to be monitored simultaneously during behavior and how the internal dynamics of the grid cell system may give rise to map changes in place cells. I will show how the collective activity of grid cells operates on a toroidal manifold, irrespective or ongoing behavior (sleep or wake) and before eyes and ears open during early postnatal development, in agreement with continuous attractor network models of grid cells. I will then show that individual modules of grid cells can under some circumstances operate independently of each other, and that when they do so, their discordant responses give rise to remapping in maps of place cells downstream in the hippocampus. I will finally demonstrate how a sweep-like firing pattern of grid cells, consisting of rapid trajectories of population firing travelling outwards from the animal´s location once per theta cycle, enables entorhinal cells to map positions well beyond the instantaneous location of a navigating animal. 

Affiliation: Harvard University

Bio: Bence Ölveczky graduated with a degree in Mechanical Engineering from the Technical University of Budapest. He worked as a journalist for a couple of years before starting his PhD in Neuroscience at Harvard University, where he studied motion processing in the retina with Markus Meister. He was a Junior Fellow in the Harvard Society of Fellows in 2004 working on birdsong learning with Michale Fee at MIT. He is currently a Professor in the Department of Organismic and Evolutionary Biology and the Center for Brain Science at Harvard University. His lab studies the neural circuit mechanisms that underlie different aspects of motor skill learning and execution, using rats as a model system. He has received several awards, including the McKnight and Klingenstein Fellowships. He lives in Cambridge, Massachusetts, with his wife Daniele and their three  children, Oscar, Eva, and Camilla.

Talk Title: Neural circuits underlying learned motor sequence execution

Abstract: Our ability to sequence movements and actions in response to unpredictable environmental events underlies our rich and adaptive behavioral repertoire. Such flexible behaviors contrast with overtrained, or automatic, motor sequences directed at specific tasks and executed the same way every time. We probed how neural circuits underlie these distinct forms of motor sequence execution by training rats on a ‘piano task’ in which the same motor sequence can be generated in response to unpredictable cues or overtrained to the point of automaticity. By measuring and manipulating neural activity in motor cortex and sensorimotor striatum, we delineate the logic by which these circuits combine to generate both flexible and automatic motor sequences. 

Bio: Agostina Palmigiano is a Simons BTI Fellow at the Center for Theoretical Neuroscience at Columbia University and an incoming Lecturer at the Gatsby Computational Neuroscience Unit (UCL). She studied Physics in the University of Buenos Aires, and was a graduate student at the Max Planck Institute for Dynamics and Self-Organization. Her research involves the development of theoretical approaches to dissect the mechanisms employed by neuronal circuits to represent and transform sensorial information, and how those mechanisms are shaped by and in conjunction with behavior.

Talk Title: Mechanisms underlying reshuffling of visual responses by optogenetic stimulation in mice and monkeys

Abstract: The ability to optogenetically perturb neural circuits opens an unprecedented window into mechanisms governing circuit function. We analyzed and theoretically modeled neuronal responses to visual and optogenetic inputs in mouse and monkey V1. In both species, optogenetic stimulation of excitatory neurons strongly modulated the activity of single neurons, yet had weak or no effects on the distribution of firing rates across the population. Thus, the optogenetic inputs reshuffled firing rates across the network. Key statistics of mouse and monkey responses lay on a continuum, with mice/monkeys occupying the low/high rate regions, respectively. We show that neuronal reshuffling emerges generically in randomly connected excitatory/inhibitory networks, provided the coupling strength (combination of recurrent coupling and external input) is sufficient that powerful inhibitory feedback cancels the mean optogenetic input. A more realistic model, distinguishing tuned visual vs.\ untuned optogenetic input in a structured network, reduces the coupling strength needed to explain reshuffling.

Affiliation: University of Geneva

Bio: Alexandre Pouget, PhD, is a full Professor at the University of Geneva in the Basis Neuroscience department. He received his undergraduate education at the Ecole Normale Supérieure, Paris, before moving to the Salk Institute in 1988 to pursue a PhD. in computational neuroscience in the Sejnowski laboratory. After a postdoc at UCLA with John Schlag in 1994, he became a professor at Georgetown University in 1996, then at the university of Rochester in the Brain and Cognitive Science department in 1999 before moving to the university of Geneva in 2011. Dr. Pouget was awarded the Carnegie Prize in Brain and Mind Sciences in 2016. He is the author of more than 100 papers and the editor of one book. He co-founded the Computational System Neuroscience conference in 2004 with Tony Zador. In 2016, he co-founded the International Brain Laboratory with Zach Mainen and Mike Hausser, the first international CERN-like collaboration in circuit neuroscience.

Talk Title: Brain wide representations of prior information in mouse decision-making

Abstract: The neural representations of prior information about the state of the world are poorly understood. To investigate this issue, we examined brain-wide Neuropixels recordings and widefield calcium imaging collected by the International Brain Laboratory. Mice were trained to indicate the location of a visual grating stimulus, which appeared on the left or right with prior probability alternating between 0.2 and 0.8 in blocks of variable length. We report that this subjective prior is encoded in at least 20% to 30% of brain regions which, remarkably, span all levels of processing, from early sensory areas (LGd, VISp) to motor regions (MOs, MOp, GRN) and high level cortical regions (ACCd, ORBvl). This widespread representation of the prior is consistent with a neural model of Bayesian inference involving loops between areas, as opposed to a model in which the prior is incorporated only in decision making areas. This study offers the first brain-wide perspective on prior encoding at cellular resolution, underscoring the importance of using large scale recordings on a single standardized task.

Affiliation: The Rockefeller University

Bio: Vanessa Ruta’s lab aims to elucidate how nervous systems are shaped by either individual experience or evolutionary selection to give rise to flexible variations in behavior.  Her group leverages the concise chemosensory circuits of Drosophila to gain an integrative understanding of adaptive behaviors, from the structure of sensory receptors to the algorithms of complex motor behavior.   Dr. Ruta received a B.A. in chemistry from Hunter College of CUNY and a Ph.D. from The Rockefeller University.  She carried out postdoctoral studies with Richard Axel at Columbia University before starting her own group at The Rockefeller University, where she is currently the Gabrielle H. Reem and Herbert J. Kayden Professor.  Dr. Ruta was named a MacArthur Fellow, a McKnight Scholar, a Pew Biomedical Scholar, and a New York Stem Cell Foundation–Robertson Neuroscience Investigator and appointed as a Howard Hughes Medical Institute Investigator in 2021.

Talk Title: Adaptive circuit algorithms for olfactory navigation

Abstract: Olfactory systems are continuously barraged with odors, varying in their structure, physicochemical properties, and concentration. Contending with such a complex chemical landscape poses a distinct challenge to the fundamental flexibility of the nervous system. Our lab has been using olfaction as a window into the mechanisms of flexible and adaptive behavior, leveraging the concise olfactory circuits of Drosophila to reveal how animals can detect, perceive, and navigate the vast chemical world.  In recent work, we developed a novel closed-loop olfactory paradigm that allows head-fixed Drosophila to navigate arbitrary chemical landscapes, shedding light on the basic neural and behavioral algorithms of olfactory plume navigation.  By combining functional imaging, precise perturbations of neural activity and behavioral modeling, we show that plume navigation represents a spatial navigation task in which animals integrate memories of their past olfactory encounters with their current sense of space in order to shape their navigational strategy. 

Affiliation: Columbia University

Bio: I performed my graduate work at Columbia University under the supervision of Dr. Randy Bruno. My PhD thesis focused on how thalamic and cortical synaptic inputs are integrated by neurons in the barrel cortex of the rat. During my postdoc in Dr. Richard Axel’s lab I have been studying the plasticity of the mouse olfactory cortex. I am particularly interested in how this circuit continuously adapts to the recent statistics of the olfactory environment.

Talk Title: Experience dependence of connectivity in olfactory cortex

Abstract: Our understanding of how experience shapes the synaptic connections of a network has been impeded by the difficulty of estimating connectivity in a living animal. We have taken advantage of features particular to the mouse olfactory cortex (piriform) to largely overcome this challenge and infer excitatory connections from spike-time cross correlograms in vivo. The piriform is a sparsely connected circuit (~0.1% connection probability) and its cytoarchitecture permits simultaneous recording of the spiking activity of order 10^3 single units within a single network. These properties, combined with long recording sessions, permit us to detect putative synapses with low error rates. For this, we developed a deep neural network approach that identifies ~10^3 connections amongst the ~10^6 possible interactions per recording. We found a positive, monotonic relationship between how similarly two neurons respond to odors and how likely they are to be connected to each other. We then asked whether prior experience with odorants alters the connectivity of the network. After experience the relationship between connection probability and the similarity of responses to familiar odors was U-shaped: anti-correlated pairs were nearly as likely to be connected as correlated pairs. Moreover, experience resulted in a drop in connections amongst neurons whose responses to familiar odors were uncorrelated. These phenomena are captured by a model in which connections between uncorrelated pairs are selectively ablated, while new connections are added independently of their neurons’ tuning.

Affiliation: University of California, Santa Cruz

Bio: Dan received his B.S in Applied Physics and Electrical Engineering from Yale University in 2008, and his Ph.D in Applied Physics from Caltech in 2013. Though Dan studied solar cells in graduate school, he switched to neuroscience for his postdoctoral training and joined the lab of Dr. Vivek Jayaraman at the HHMI Janelia Research Campus. At Janelia, Dan leveraged a connectome of the Drosophila brain to identify the circuit motifs that shape information as it flows through the fly navigation brain region. He then compared this connectivity data to theoretical models of ring attractor networks to determine the mechanisms that underlie the fly head direction system and used a wide range of experimental approaches to demonstrate how these mechanisms maintain and update the head direction representation. Dan joined the MCD Biology faculty at the University of California, Santa Cruz in 2021, where he leads the Ground Up Neuroscience Group. His group works to develop a bottom-up approach to determine how neural functions emerge from neural connectivity and cellular biophysics.

Talk Title: Maintaining, updating, and relaying a head direction representation in the fruit fly brain

Abstract: Internal representations of head direction can arise from ring attractor networks. These networks feature different cell types with distinct roles in maintaining, updating, and relaying the head direction signal. Neurons within and across these cell types are predicted to have rotationally invariant connectivity profiles with connections that are excitatory, inhibitory, or neuromodulatory depending on their functional role. We tested these predictions by analyzing the recent connectomes of the Drosophila central brain, performing cell type specific RNA-Seq and FISH, tagging neurotransmitter receptor proteins, and monitoring changes in the head direction representation in response to cell type specific perturbations. Our results largely confirmed long standing ring attractor models while revealing unpredicted cell types necessary for maintaining or updating the head direction representation and unpredicted mechanisms for relaying the head direction representation to downstream circuits. We will discuss these novel elements and their effects on the dynamics of head direction activity in the Drosophila ring attractor network. 

Affiliation: Department of Brain Sciences, Weizmann Institute of Science, Rehovot, Israel

Bio: Nachum Ulanovsky is a Professor of Neuroscience and Head of the Center for Learning, Memory & Cognition at the Weizmann Institute of Science. He is an elected International Honorary Member of the American Academy of Arts and Sciences. Nachum studies the neural basis of spatial cognition and social cognition in the mammalian hippocampal formation – using bats as a novel animal model that he pioneered. His group developed wireless-electrophysiology devices, which enabled the discovery of 3D place-cells, 3D head-direction cells and 3D grid cells in flying bats; as well as the discovery of goal-vector cells – which encode navigational goals; and social place-cells – which represent other individuals, in a social context. Nachum performs experiments in bats flying in very large environments, hundreds of meters in size, during navigation and social interactions. He seeks to create a “Natural Neuroscience” approach for studying the neural basis of behavior – tapping into the animal’s natural behaviors in complex, large-scale, naturalistic settings.

Talk Title: Neural codes for natural behaviors in flying bats

Abstract: This talk will focus on the importance of using natural behaviors in neuroscience research – the “Natural Neuroscience” approach. I will illustrate this point by describing studies of neural codes for spatial behaviors and social behaviors, in flying bats – using wireless neurophysiology methods that we developed – and will highlight new neuronal representations that we discovered in animals navigating through 3D spaces, or in very large-scale environments.  In particular, I will discuss three recent studies: (1) A multi-scale neural code for very large environments, which we discovered in the hippocampus of bats flying in a 200-meter long tunnel. This new type of neural code is fundamentally different from spatial codes reported in small environments – and we show theoretically that it is superior for representing very large spaces.  (2) Rapid modulation of position × distance coding in the hippocampus during collision-avoidance behavior between two bats flying in the long tunnel. This result provides a dramatic illustration of the extreme dynamism of the neural code.  (3) I will also describe new results on the social representation of other individuals in the hippocampus, in a highly social multi-animal setting – which revealed complex neuronal coding of social variables.   The lecture will propose that neuroscience experiments – in bats, rodents, monkeys or humans – should be conducted under evermore naturalistic conditions. 

Affiliation: Institute of Neuroscience (ION), Chinese Academy of Sciences

Bio: Dr. Liping Wang studied tactile working memory in non-human primates under the supervision of Profs. Yong-Di Zhou and Steven Hsiao at the Johns Hopkins University during his Ph.D. (2004-2009). He then did his postdoctoral work with Prof. Yasushi Miyashita at the University of Tokyo in Japan (2010-2011) and Prof. Stanislas Dehaene at the INSERM
NeuroSpin brain imaging center in France (2012-2015). In Paris, Wang focused on neural representations of language syntax in both humans and non-human primates. In 2016, Wang joined the Institute of Neuroscience in the Chinese Academy of Sciences. He is now a senior Investigator and Head of the Laboratory of Cognition. His main interests are the neural mechanisms underlying sequence learning, working memory, and consciousness.

Talk Title: The control of sequence working memory in macaque frontal cortex

Abstract: To memorize a sequence, one must serially bind each item to its order. In this process, how the brain flexibly controls a given input to its associated order in sequence working memory (SWM) remains unknown. Here, we investigated neural representations of the SWM control using high-throughput electrophysiological recordings to record thousands of neurons in the frontal cortex of macaque monkeys performing delayed sequence-sorting tasks. We found separate representations of sensory and SWM at the population level within the same frontal circuitry, reflecting the control process through their organized neural dynamics. Item at each rank was sequentially entered into a common entry subspace and, depending on the sorting algorithm (forward or backward), flexibly and timely controlled into their associated rank-selective SWM subspaces. Crucially, neural activities in these subspaces can faithfully predict sequence recall performance and errors in single trials. Thus, well orchestrated neural dynamics in the frontal cortex underlie the flexible control of SWM.

Affiliation: Princeton University

Bio: Ilana Witten is a professor of neuroscience at Princeton University. She was first introduced to neuroscience research as an undergraduate at Princeton, where she majored in physics. She then obtained a PhD from Stanford University in neuroscience in 2008, where she worked in the systems neuroscience lab of Eric Knudsen and collaborated with Haim Sompolinsky of Hebrew University. As a postdoctoral fellow, she worked with Karl Deisseroth at Stanford University, and has been on the faculty at Princeton since 2012.  By developing and applying new experimental technologies, and integrating them with a computational perspective, her research program has broken ground in the understanding of how learning and decision-making algorithms map onto brain circuitry. She has received multiple awards for her work, including an NIH New Innovator Award, a Mcknight Scholars Award, a NYSCF Investigator Award, and the Daniel X Freedman Prize.

Talk Title: Learning from post-ingestive feedback

Abstract: Animals learn the value of foods based on their postingestive effects. However, it remains unclear how the brain is able to associate flavors experienced during a meal with visceral feedback signals that arise after a delay of minutes or hours. To gain insight into this problem, we leveraged the fact that mice learn to associate novel — but not familiar — foods with gastric malaise signals, and investigated what distinguishes the neural representations of flavors that promote learning versus those that do not. Surveying brainwide expression of the immediate early gene Fos during drinking and postingestive malaise revealed that a cluster of amygdala regions, especially the central amygdala (CEA), are preferentially activated when mice consume novel flavors that permit learning and that this activation is amplified by the arrival of malaise signals from the gut. This suggests that a dedicated CEA ensemble may directly link flavors to delayed visceral malaise signals. To test this hypothesis, we used chronic Neuropixels recordings to track the activity of individual CEA units across drinking and malaise. This revealed that malaise signals preferentially reactivate novel flavor-coding units versus other CEA cell types, and that population dynamics during bouts of malaise closely mirror those during novel flavor drinking. Moreover, neurons that were most activated by malaise feedback showed increased activity during a subsequent retrieval test. Together, our findings suggest that reactivation of relevant flavor representations provides a mechanism to bridge the gap between drinking and delayed feedback and thereby associate flavors with their postingestive consequences. 

Affiliation: Stanford University

Affiliation: Cold Spring Harbor Laboratory

Bio: Anthony Zador received his MD and PhD from Yale in 1994, where his focus was theoretical neuroscience and neural networks.  He then did postdoctoral research in synaptic physiology at the Salk Institute in La Jolla, California.  In 1999, he joined the faculty at Cold Spring Harbor Laboratory in New York, where he is now the Alle Davis Harris Professor of Biology and served as Chair of Neuroscience from 2008-2018.  The goal of his research is to understand the neural circuits underlying cortical processing.  His laboratory pioneered the use of rodents in complex sensory decision tasks, and also developed a revolutionary approach to determining brain wiring using high-throughput DNA sequencing.  His current research interests include applying neuroscience to usher in the next generation of Artificial Intelligence.

Talk Title: From Circuitry to Computation

Abstract: How do diverse cell types interact within the brain to drive complex behavior and support specific computations? Understanding these interactions, and the interplay between unique functional properties, transcriptomes, and connectivity patterns, can lead to a more comprehensive view of neural circuits and their role in behavior. I will discuss our work on combining in vivo recording with in vitro characterization, including in situ transcriptomics and barcode-based projection mapping, to systematically relate circuitry and activity.