The Satellite Workshops of the Bernstein Conference provide a stage to discuss topical research questions, novel scientific approaches, and challenges in computational neuroscience and related fields.
All workshops will be held in parallel sessions and moderated by the workshop organizers.
Monday, Sep 28, 14:00 – 18:30 CEST
Tuesday, Sep 29, 8:30 – 12:30 CEST
Goethe University
Campus Westend / Seminarhaus
Max-Horkheimer-Straße 4
60323 Frankfurt am Main
Germany
Katharina Wilmes | Institute of Neuroinformatics in Zurich, Switzerland (Chair)
Juan Álvaro Gallego | Champalimaud Foundation, Portugal (Workshop Vice Chair)
Plasticity and function of hippocampal memory circuits
Organizers:
Samuel Eckmann | University of Cambridge, UK
Uri Cohen | Weizmann Institute of Science, Israel
Description
This workshop aims to bridge the gap between theoretical models and experimental realities of hippocampal memory processing. We will explore how theoretical models of high-capacity associative memory may be supported by dendritic processing and the detailed connectivity between specific neuron subtypes. We further highlight the tension between one-shot memory storage via Behavioral Timescale Synaptic Plasticity (BTSP) and incremental STDP models in light of the emerging role of memory replay and consolidation. In summary, our workshop aims to facilitate a shift from the hippocampus as a static storage center to a dynamic memory system capable of generalizing across diverse contexts.
Schedule: Monday, Sept 28
| 14:00 | Dmitry Krotov (Independent Researcher, USA) |
Dense Associative Memory and its potential role in brain computation |
| 14:30 | Mohadeseh Shafiei Kafraj (University College London, UK) |
A dendritic associative memory with compositional representations |
| 15:00 | Everton Agnes (University of Basel, Switzerland) |
Attractors meet dendrites: a circuit mechanism for selective recall in high-capacity memory networks |
| 15:30 | Jake Watson (Hospital del Mar Research Institute, Spain) |
From cell types to circuits: hippocampal CA3 organisation across species |
| 16:00 | Coffee break | |
| 16:30 | Jozsef Csicsvari (Institute of Science and Technology Austria (ISTA), Austria) |
Converging manifold coding in the hippocampus and medial prefrontal cortex across increasing experience in spatial memory tasks |
| 17:00 | Judit Makara (HUN-REN Institute of Experimental Medicine, Hungary) |
Reorganisation of hippocampal representations by learning and changing contexts |
| 17:30 | Henning Sprekeler (Technische Universität Berlin, Germany) |
Memory consolidation and representational drift |
| 18:00 | Discussion | |
How the parts work together as a whole: From brain-wide neural representations to computations
Organizers:
Douglas Feitosa Tomé | Institute of Science and Technology Austria (ISTA), Austria
Adrienne Fairhall | University of Washington, USA
Description
Recent technological advances have enabled large-scale recordings across the brain, revealing distributed neural representations that span multiple regions and support perception, cognition, and behavior. While these developments have expanded our ability to map brain-wide dynamics, they raise a fundamental unresolved question: how are distributed neural representations coordinated to implement computations? This workshop will address this question by bringing together cutting-edge work at the interface of theoretical and experimental neuroscience. Bridging these domains is essential: the scale and complexity of brain-wide recordings require new theoretical frameworks to uncover underlying computations, while emerging theories of distributed computation demand targeted experimental validation. The program will include an introductory overview, research talks combining data-driven modeling with theory-guided experiments, and dedicated discussion sessions to foster active engagement and critical debate. We will focus on key questions such as how computations are distributed across brain regions, whether trade-offs exist between localized and distributed processing, and how inter-regional interactions give rise to coordinated function. By integrating perspectives across species, recording techniques, and computational frameworks, the workshop aims to synthesize current advances and identify general principles governing distributed neural computation.
Schedule: Monday, Sept 28
| 14:00 | Douglas Feitosa Tomé (Institute of Science and Technology Austria (ISTA), Austria) |
Introduction |
| 14:05 | Tatiana Engel (Princeton University, USA) |
Brain-wide organization of intrinsic timescales at single-neuron resolution |
| 14:30 | Charles Findling (University of Geneva, Switzerland) |
Brain-wide representations of prior information in mouse decision-making |
| 14:55 | Denis Alevi (Technische Universität Berlin, Germany) |
Memory consolidation and representational drift |
| 15:20 | Gabriel Kreiman (Harvard Medical School, USA) |
What AI wants to be when it grows up: a brain |
| 15:45 | Discussion with all speakers | |
| 16:00 | Coffee break | |
| 16:30 | Adrienne Fairhall (University of Washington, USA) |
Distributed neural representations of beliefs mediate probabilistic inference |
| 16:55 | Douglas Feitosa Tomé (Institute of Science and Technology Austria (ISTA), Austria) |
Distributed engrams enable parallelized orthogonal computations within and across brain regions |
| 17:20 | Jennifer Li (Max Planck Institute for Biological Cybernetics, Germany) |
Brain-wide joint analysis of neuromodulatory and cognitive networks during spatial learning in freely swimming zebrafish |
| 17:45 | Claudia Clopath (Imperial College London, UK) |
Why motor learning involves multiple systems: an algorithmic perspective |
| 18:10 | Discussion with all speakers | |
Neuromodulatory computations across the lifespan
Organizers:
Srikanth Ramaswamy | Newcastle University, UK
Trang-Anh Nghiem | Hertie Institute of AI in Brain Health, Germany
Description
Neuromodulation plays a key role in shaping neural computations and enabling adaptable learning and behavior throughout the lifetime. Accordingly, disrupted neuromodulation is hypothesized to underlie pathological development and aging: pharmacological interventions targeting neuromodulators such as dopamine, acetylcholine, and serotonin have been widely leveraged in neurodevelopmental disorders like attention deficit hyperactivity disorder and schizophrenia, as well as neurodegenerative diseases like Parkinson’s and Alzheimer’s. Yet, the mechanisms by which neuromodulators control cellular and synaptic mechanisms, influence neural dynamics, and modulate encoding and learning across the lifespan remains largely unknown. In this workshop, we aim to discuss how neuromodulatory systems, their associated neural representations, and learning rules change during development, aging, and related disorders. To this end, we will bring together theoretical and experimental experts on neuromodulation, neural computations, as well as neurodevelopment, aging, and their relevant pathologies. Our workshop will foster collaborations and generate actionable research questions for future multi-scale, lifetime-spanning studies. This is timely given the urgency posed by aging populations, rising neurodevelopmental diagnoses, and advances in neuroAI architectures leveraging neuromodulation-like mechanisms. By addressing mechanistic links between development, aging, and pathology, the workshop will catalyse innovations in diagnosis, therapeutic strategies, and machine learning systems. Ultimately, participants will identify critical knowledge gaps, develop interdisciplinary approaches, and lay the groundwork for transformative discoveries in brain health and adaptive intelligence across the human lifespan.
Schedule: Monday, Sept 28
| 14:00 | Shervin Safavi (TUD Dresden University of Technology, Germany) |
Criticality, neuromodulatory state, and the computational machinery of perceptual decision-making |
| 14:40 | Julia Costacurta (Stanford University, USA) |
Statistical signatures of neuromodulatory state in neural population dynamics |
| 15:10 | Suraj Honnuraiah (Harvard Medical School, USA) |
Dendritic computation and neuromodulatory gain control: From neural circuits to neuromorphic architectures |
| 16:00 | Coffee break | |
| 16:30 | Robert Froemke (New York University, USA) |
Oxytocin as a neuromodulatory gatekeeper: Synaptic plasticity, social learning, and lifespan implications |
| 17:15 | Fani Koulkouli (Institut national de la santé et de la recherche médicale (INSERM), France) |
Nicotinic acetylcholine receptors, cortical interneurons, and the breakdown of circuit balance in disease |
| 18:00 | Ylermi Cabrera Leon (Universidad de Las Palmas de Gran Canaria, Spain) |
Sleep, spontaneous dynamics, and the neuromodulatory regulation of memory consolidation |
Towards a quantitative approach to behavior
Organizers:
Pietro Verzelli | Bonn University Hospital, Germany
Jens Tillmann | Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE), Germany
Description
Behavior is the primary output of neural systems and a central object of study in neuroscience. Recent advances in experimental techniques and machine learning have made it possible to collect large-scale, high-resolution behavioral data across species and contexts. There is broad agreement that understanding behavior is essential for linking neural activity to function. Despite this abundance of data, there is still no shared framework for how to represent, structure, or interpret behavior. Moreover, at a conceptual level, even the question of what should count as behavior remains open: if one were to gather researchers and ask “what is behavior?”, one would likely obtain a wide range of answers. This is precisely what this workshop is about. We bring together researchers working on different model organisms and using complementary approaches—experimental, computational, and theoretical—to address how structure can be identified and formalized in behavior. Topics will include low-dimensional representations, data-driven embeddings, and the integration of behavioral and neural measurements. The workshop will place particular emphasis on comparing perspectives: what different communities consider as behavior, how they represent it, and what can be learned by confronting these viewpoints. By doing so, the goal is not only to provide a clearer picture of the current landscape, but also to take a step towards a more unified and principled quantitative science of behavior.
Schedule: Monday, Sept 28
| 14:00 | Jens Tillmann (Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE), Germany) |
Introduction: What do we talk about when we talk about behavior? |
| 14:30 | Ahmed Al-Hady & Jacob Davidson (Max Planck Institute of Animal Behavior, Germany) |
Social behavioral analytics for natural neuroscience |
| 15:00 | Graziana Gatto (University Hospital Cologne, Germany) |
Behavior in motion: Standardizing how we measure action |
| 15:30 | Fabio Nächt (University of Vienna, Austria) |
From high-resolution behavior quantification to predictive optogenetic virtual realities in C. elegans |
| 16:00 | Coffee break | |
| 16:30 | Dominik Bach (University of Bonn, Germany) |
The grammar of human behaviour in a biological environment |
| 17:00 | Mostafa Safaie (Imperial College, UK) |
Control of striatal activity reflects different circuit constraints compared to M1 |
| 17:30 | Fabienne Rössler (ETH Zurich, Switzerland) |
Behavioral profiling to predict individual stress vulnerability |
| 18:00 | Pietro Verzelli (Bonn University Hospital, Germany) |
Closing discussion: Towards a quantitative science of behavior |
Controversies on the nature of spatial coding in the hippocampal formation across species
Organizers:
Andrej Bicanski | Max Planck Institute for Human Cognitive and Brain Sciences, Germany
Josh Jacobs | University of Chicago, USA
Richard Kempter | Humboldt-Universität zu Berlin, Germany
Lukas Kunz | University of Bonn, Germany
Description
The hippocampal formation houses interpretable neural codes such as place cells, grid cells, head direction cells, and vectorial representations, long thought to underpin spatial navigation and memory. However, despite decades of research, fundamental questions about the nature and generality of these representations remain actively contested. This workshop brings together leading and early-career scientists to debate, stress-test, and where possible advance resolution on several key controversies in the field. A central theme is whether the spatial coding principles established in rodents generalize across species. Findings in non-human primates, and accumulating data from humans, suggest that the hippocampal system may operate differently across mammalian lineages. At the same time, select invasive, and numerous non-invasive studies suggest commonalities. Thus apparent differences could also reflect experimental and behavioral factors rather than distinct computational architectures. Broader questions include the nature of representational drift, the relationship between innate and learned spatial representations, and whether the human hippocampus deploys fundamentally new computational strategies in complex environments. Rather than seeking premature consensus, the workshop embraces productive disagreement as a scientific method, followed by extended open discussion. We prioritize exchange over exposition with the goal to sharpen questions that will drive the field forward.
Schedule: Tuesday, Sept 29
| Block 1 | What do human MTL neurons actually encode? | Chair: Andrej Bicanski |
| 8:00 | Josh Jacobs (University of Chicago, USA) |
Introduction and neural coding in the human hippocampus during navigation |
| 8:20 | Lukas Kunz (University of Bonn, Germany) |
Investigating the role of concept cells in human goal-directed navigation |
| 8:40 | Open discussion: Block 1 | |
| Block 2 | Human behavior vs. human neural representation | Chair: Josh Jacobs |
| 9:00 | Shachar Maidenbaum (Ben-Gurion University of the Negev, Israel) |
Spatial representations for memory and navigation across the reality spectrum |
| 9:20 | Ikwan Bin Khalid / Richard Kempter (Humboldt-Universität zu Berlin, Germany) |
Emergence of macroscopic grid-like representations |
| 9:40 | Open discussion: Block 2 | |
| 10:00 | Coffee break | |
| Block 3 | Drift vs. structure | Chair: Lukas Kunz |
| 10:30 | Sang Ah Lee (Seoul National University, South Korea) |
Neural mechanisms of spatiotemporal binding in navigation and memory |
| 10:50 | Natalie Schieferstein (University of Bonn, Germany) |
Representational drift: Generating mechanisms and implications for neural coding |
| 11:10 | Open discussion: Block 3 | |
| Block 4 | Same neural machinery, many niches? | Chair: Sang Ah Lee |
| 11:30 | Dun Mao (Chinese Academy of Sciences, China) |
Alternative spatial representations in monkeys |
| 11:50 | Liora Las (Weizmann Institute of Science, Israel) |
Unraveling new hippocampal-entorhinal coding schemes in bats through natural behaviors |
| 12:10 | Andrej Bicanski (Max Planck Institute for Human Cognitive and Brain Sciences, Germany) |
Repurposing extant spatial navigation architectures across species and cognitive domains |
| Open discussion: Block 4 | ||
Degeneracy in nervous system function
Organizers:
Peter Jedlicka | Goethe University Frankfurt, Germany
Luisa Ramirez | Johannes Gutenberg University Mainz, Germany
Description
The brain integrates an astonishing amount of information to guide appropriate functional outcomes. To achieve this, while accounting for a wide range of internal and external conditions, neuronal function must be both reliable and versatile. Degeneracy — the ability of structurally different elements to produce equivalent functional outcomes — is emerging as a fundamental mechanism that enables this. Evidence for degeneracy exists at every level of nervous system organization: in the many ion channel combinations that yield the same neuronal firing properties, in the parallel circuit configurations that support identical computations, and in the distinct neuronal ensembles that encode the same stimulus. Yet despite this ubiquity, degeneracy has been called the “brain’s best kept secret” (Albantakis et al., 2024, J. Neurosci.), and remains understudied as an organizing principle of neural function. This workshop brings together experimental and computational neuroscientists who have directly demonstrated degeneracy in their systems, spanning multiple levels of organization — from ion channels to neural populations — and across invertebrate and vertebrate species. By combining empirical and theoretical perspectives, we aim to establish degeneracy as a unifying framework for understanding how the nervous system achieves both robustness and versatility, and to stimulate discussion on its mechanistic origins, functional consequences, and normative underpinnings. We will open the workshop at the level of single neurons. Dr. Stefanie Ryglewski (JGU Mainz) will present experimental evidence that heterogeneity in ion channel isoforms safeguards neuronal excitability in flight motoneurons, building on a rich framework of computational predictions. Her talk will start the discussion on how structural variability, far from being noise, is a feature that the nervous system exploits to ensure functional robustness. To continue this discussion from a theoretical angle, we will ask how heterogeneity at the cellular level shapes the dynamics and computational properties of entire networks. Dr. Richard Gast (The Scripps Research Institute, USA) will address this directly by presenting mean-field frameworks for neural populations with heterogeneous spiking properties and discussing how within-type variability controls the repertoire of computations a network can perform. Zooming into a concrete circuit, Dr. Juan Vargas (JGU Mainz) will then present striking experimental evidence of degeneracy in the Drosophila motion detection pathway. He will show that direction-selective responses can be recovered by partial circuit rescues, demonstrating that structurally distinct configurations can achieve the same computation. This talk will provide experimental evidence of circuit-level degeneracy, raising the question of whether such degeneracy is particular of the fly visual system, or a more general organizational principle. Prof. Fleur Zeldenrust (Donders Institute for Brain, Cognition and Behaviour, The Netherlands) will argue for the latter. From a theoretical perspective, she will show that heterogeneous recurrent networks can achieve equivalent and robust information transfer across a wide range of structural configurations. She will discuss the energy-information trade-offs that shape which degenerate solutions the nervous system selects — connecting the biophysical level back to normative principles. We will next shift to vertebrate systems, where degeneracy manifests at the level of neural population codes. Prof. Simon Rumpel (Johannes Gutenberg University, Mainz) will show that in the mouse auditory cortex, the same stable representational map can be realized by a large number of distinct neuronal configurations. Remarkably, this degeneracy is not merely a static property: it can be directly observed over time, as sound-evoked activity patterns drift across days in the same individual — a phenomenon known as representational drift — while perception remains unchanged. Building on this, Prof. Stephanie Palmer (University of Chicago, USA) will discuss the principles behind the different degenerate solutions in nervous systems. Drawing on her work on predictive coding in the retina and visual cortex, she will show that neural populations are optimized to predict their future inputs — and that this predictive optimality provides a powerful normative framework for understanding why certain degenerate configurations are favored over others. Prof. Elad Schneidman (Weizmann Institute of Science, Israel) will close the workshop by asking what degeneracy means from a principled, information-theoretic standpoint. Using maximum entropy models and tools from statistical physics, he will address what is preserved and what varies across degenerate population codes, and how the architecture of neural circuits — from local connectivity to connectome-level wiring — shapes the space of equivalent solutions available to the brain. Together, our symposium shows many intriguing examples of degeneracy in nervous system function. Across all talks, we will identify — both through experimental and theoretical approaches — how degeneracy serves as a ‘built-in’ mechanism that is immediately available at many levels in the nervous system to ensure reliable and versatile function. At the same time, we will discuss the limitations and trade-offs of such a system. With this, we aim to promote this understudied topic as a useful, if not necessary, framework for understanding nervous system function.
Schedule: Tuesday, Sept 29
| 8:30 | Stefanie Ryglewski (Johannes Gutenberg University, Germany) |
Robustness through variability: ion channel isoform diversity safeguards neuronal excitability |
| 9:00 | Alexander Bird (Justus Liebig University Giessen, Germany) |
Degeneracy and optimality in single cell computations |
| 9:30 | Fleur Zeldenrust (Donders Institute for Brain, Cognition, and Behaviour, The Netherlands) |
Heterogeneity in the brain: Lessons from theory and experiment |
| 10:00 | Simon Rumpel (Johannes Gutenberg University Mainz, Germany) |
Degenerate coding of sensory stimuli in the neocortex |
| 10:30 | Coffee break | |
| 11:00 | Matthew Chalk (Sorbonne University, France) |
From mutual information to geometry: characterising degeneracy in neural population codes |
| 11:30 | Juan Vargas (Johannes Gutenberg University Mainz, Germany) |
Coexisting implementations of direction selectivity in the Drosophila visual system |
| 12:00 | Elad Schneidman (Weizmann Institute of Science, Israel) |
Architectural design principles that shape the structure and computations of neural circuits |
| 12:30 | Discussion and Questions | |
Exotic extracellular waveforms: can we separate non-somatic signals from artefacts?
Organizers:
Paula Kuokkanen | Humboldt-Universität zu Berlin, Germany
Jérémie Sibille | Charité – Universitätsmedizin Berlin, Germany
Description
The extracellular neuronal action potentials show a multitude of waveforms beyond the best-known somatic spikes. Detecting and sorting these shapes remains challenging in neuroscience, especially given the wealth of newer observations stemming from high-density silicon probes. Furthermore, a substantial fraction of the more exotic waveforms in electrophysiological recordings can be missed by current detection algorithms due to the uncertainty about their origin, rendering parts of the material unusable. Aiming for generalizable solutions, this workshop will approach this question of the origin of exotic extracellular waveforms by shedding new light based on putative contributions from dendrites, soma, and axons. As a matter of fact, biophysics has already predicted that each of these compartments can generate distinct extracellular waveforms, which do contribute to extracellular activity; their presence in the actual recorded spikes, by, e.g., Neuropixels, still remains a limit more than an improvement. The central aim of this workshop is to build a landscape of the expected waveform from these three compartments, given their known biophysical properties, toward more inclusive, and hopefully more precise, interpretations of in vivo electrophysiological data.
Schedule: Tuesday, Sept 29
| 8:30 | Jérémie Sibille (Charité – Universitätsmedizin Berlin, Germany) |
The bestiary of extracellular waveforms from different neuron compartments |
| 8:45 | Gaute Einevoll (University of Oslo & Norwegian University of Life Sciences (NBMU), Norway) |
Biophysics of extracellular waveforms and their summation |
| 9:15 | Ian Christopher Tanoh (Stanford University, USA) |
Identifying multi-compartment Hodgkin-Huxley models with high-density extracellular voltage recordings |
| 9:35 | Rishikesh Narayanan (Indian Institute of Science, India) |
Active dendritic and gap junctional contributions to extracellular field potentials |
| 10:00 | Coffee break | |
| 10:30 | Alexandra Tzilivaki (Charité – Universitätsmedizin Berlin, Germany) |
From cellular mechanisms to network connectivity: GABAergic interneurons shape memory-related oscillations |
| 10:50 | Simon Musall (Forschungszentrum Jülich, Germany) |
UnitRefine: A Community Toolbox for Automated Spike Sorting Curation |
| 11:15 | Angelique Paulck (Harvard Medical School & Massachusetts General Research Institute, USA) |
Extracellular waveforms in human recordings |
| 11:40 | Eran Stark (University of Haifa, Israel) |
Positive, biphasic, and triphasic extracellular waveforms correspond to return currents and non-somatic spikes |
| 12:10 | Extended discussion | |
Dendritic inhibition for efficient computation and learning
Organizers:
Lucas Rudelt | Max Planck Institute for Dynamics and Self-Organization, Germany
Viola Priesemann | Max Planck Institute for Dynamics and Self-Organization, Germany
Description
Experimental and theoretical work of the last decades established the balance of excitation and inhibition to be key to support neural computations. Traditionally, this has referred to a co-tuning of excitation and inhibition at the level of neurons or populations, with substantial impact on neural network dynamics. However, recent experimental evidence suggests a coordination of excitation and inhibition on an even finer scale: Inhibition can have a strong, localized effect on individual dendrites, inhibitory plasticity depends on local postsynaptic signaling, and vice versa, inhibition has a strong modulatory effect on the induction of local excitatory plasticity. However, in contrast to E-I balance at the neuronal level, our understanding of the functional implications of dendritic inhibition and heterosynaptic plasticity at the level of individual dendrites is still fragmented. In this workshop, we therefore bring together experimentalists and theoreticians to shed light on the role of a dendritic inhibition for neural computation and learning. In particular, we will discuss (i) mechanisms of inhibitory plasticity that can establish a local co-tuning between excitation and inhibition, (ii) the implications of dendritic inhibition on subcellular and network computations, and (iii) the role of inhibition for heterosynaptic plasticity and learning. By approaching the problem from these different angles, the workshop aims at providing a more coherent picture of the intricate interplay of excitation and inhibition at dendrites, and its importance for efficient neural computation and learning.
Schedule: Tuesday, Sept 29
| 8:30 | Josef Bischofberger (University of Basel, Switzerland) |
Environmental enrichment improves learning and hippocampal sparse coding via enhanced dendritic inhibition |
| 9:00 | Corette J. Wierenga (Radboud University, The Netherlands) |
Dendritic coordination of excitatory and inhibitory synapses – a role for neuromodulation? |
| 9:30 | Eleonora Pali (Pavia University, Italy) |
Dendritic processing drives spike-timing dependent plasticity (STDP) in cerebellar Golgi cells |
| 10:00 | Coffee break | |
| 10:30 | Henning Sprekeler (Technische Universität Berlin, Germany) |
Optimizing interneuron circuits for compartment-specific feedback inhibition |
| 11:00 | Everton Joao Agnes (University of Basel, Switzerland) |
Inhibitory plasticity for local control of excitatory learning |
| 11:30 | Gaia Tavosanis (RWTH Aachen University, Germany) |
Modularity of inhibition in a cerebellum-like circuit |
| 12:00 | Albert Gidon (Humboldt Universität zu Berlin, Germany) |
The right way to inhibit a neuron |
“1, 2, 3, 4, 5, Once I Caught a Fish Alive”: Sequence learning in recurrent neural networks
Organizers:
Andrey Formozov | Heidelberg University, Germany
Mihai A. Petrovici | University of Bern, Switzerland
Martin Both | Heidelberg University, Germany
Description
Learning and representing temporal structure is a fundamental challenge for neural systems. From episodic memory and navigation to motor control and cognition, the brain must learn, store, and flexibly replay information that unfolds over time and spans multiple temporal scales, from moments to a lifetime. Recurrent neural networks (RNNs) provide a natural conceptual framework for this problem, as their intrinsic dynamics enable memory, sequence generation, and temporal integration. Yet, despite substantial progress, key questions remain unresolved: How do recurrent networks learn complex temporal structures using biologically plausible mechanisms? How are learned sequences maintained and replayed while neural representations continue to drift? These questions led to the development of a plethora of distinct approaches to building fundamental theories of brain operation and the interpretation of observable experimental data.
This workshop brings together complementary perspectives from experimental studies, theoretical neuroscience, and computational modeling to explore mechanisms of sequence learning, replay, and memory in recurrent networks. Particular emphasis lies on biologically grounded approaches, which are continuously debated (what does “biologically-plausible” truly mean?), including local and multi-factor plasticity rules, spiking dynamics, and the role of cellular and circuit-level structure in shaping temporal representations. Topics include the emergence and replay of sequences in recurrent circuits, the interaction between network dynamics and synaptic plasticity, and the stability of memory representations in the presence of drift.
The event is a gathering of true “multi-instrumentalists” with broad, mutually overlapping expertise and interests, that will tackle the problem of sequence learning from complementary perspectives. The workshop will open with a joint team talk by Rosanna Sammons and Stefano Masserini on the distinct aspects of the structure and function of the CA3 hippocampal module, with insights from anatomical and electrophysiological experimental work as well as computational modeling. From the following team talk by Icaro Lopez Costa and Tatjana Tchumatchenko, we will learn how artificial agents can learn to target multiple goals in complex environments. Mattia Chini will give his vision and overview of empirical evidence on preconfigured architectures in developing brain and emergence of protosequences, while Ben von Hünerbein will unfold the discussion on modeling of sequence learning, which, on the one hand, draws on recent advances in neurobiology and, on the other hand, holds a potential for highly efficient implementations on neuromorphic hardware. Finally, Eleanor Holton will demonstrate how different representational solutions in artificial neural networks capture individual differences in human learning and generalisation, concluding the workshop’s evening session.
The morning session of the workshop will start with a brief recap of the previous day and a talk by Andrew Lehr in which he will guide us through advances in our understanding of sequence learning, presenting circuit mechanisms that enable local dynamic control of neural sequences and how these “control knobs” can be leveraged. Next, Manuel Brenner will demonstrate an approach to dissecting the functional role of nonlinearity in recurrent networks, and Claudia Clopath will present how temporal backbones that support sequence generation can be modeled in neuronal circuits. Finally, Raoul-Martin Memmesheimer will provide us with the insights on how structured sequences can be learned and consolidated across the brain in the presence of representational drift. The session will conclude with a comprehensive recap and a vivid, integrative discussion that emphasizes the role of incorporating biological insights into computational models, while also exploring novel approaches to the validation of such models using experimental data.
Schedule: Monday, Sept 28
| 14:00 | Opening and overview of the workshop | Intro: Sequence learning in recurrent neural networks |
| 14:10 | Rosanna Sammons & Stefano Masserini (Charité – Universitätsmedizin Berlin & Humboldt Universität of Berlin, Germany) |
Structure of the CA3 pyramidal population and its influence on the local network |
| 15:05 | Icaro Lopez Costa & Tatjana Tchumatchenko (Bonn University, Germany) |
Solving a sequence of 2D navigational tasks with ANNs (to be updated) |
| 16:00 | Coffee break | |
| 16:30 | Mattia Chini (GIGA – University Liège, Belgium) |
The preconfigured brain: from synapses to circuits and behavior |
| 17:05 | Ben von Hünerbein (University of Bern, Switzerland) |
Biologically plausible learning of complex sequences in structured recurrent neural networks |
| 17:40 | Eleanor Holton (Princeton University, USA) |
Humans and neural networks show similar patterns of transfer and interference during continual learning |
| 18:15 | Recap, Q&A and discussion | |
Schedule: Tuesday, Sept 29
| 8:30 | A brief overview of the previous day | |
| 8:50 | Andrew Lehr (University Medical Center Göttingen, Germany) |
The dynamic control of neuronal activity sequences and their functional implication |
| 9:25 | Manuel Brenner (Ernst Strüngmann Institute, Germany) |
Computational roles of nonlinearity in sequence modeling |
| 10:00 | Coffee break | |
| 10:30 | Alexandra Voce (Imperial College London, UK) |
A cerebellar-inspired module accelerates sequence learning in recurrent neural networks |
| 11:05 | Raoul-Martin Memmesheimer (University of Bonn, Germany) |
Structural sequences: learning, drift and consolidation (to be updated) |
| 11:40 | Recap, Q&A and discussion | Biologically plausible models of the sequence learning and their validation |
Shaping plasticity: Structural constraints and temporal dynamics in learning
Organizers:
Anna-Maria Jürgensen | University of Cambridge & Imperial College London, UK
Emmanouil Giannakakis | Imperial College London, UK
Description
Synaptic plasticity rules operate over the signals made available by network connectivity. The anatomical and functional structure of a circuit determines which neural representations contribute to pre- and postsynaptic activity, reinforcement signals, recurrent feedback, or additional modulatory factors. As a result, learning is inherently constrained by network architecture: Connectivity defines the information accessible to plasticity rules, thereby shaping the landscape of possible associative changes. This inductive bias, imposed by network architecture, structurally partitions plasticity by determining which signals can jointly influence synaptic plasticity across spatial scales, from individual synapses to local circuits and distributed brain regions. Such partitioning increases the precision of credit assignment and reduces interference, supporting continual learning and stabilizing previously acquired associations. Additionally, these structural constraints allow learning based on weaker or noisier representations and enable generalization by preconfiguring circuits to preferentially link particular representations. The temporal dynamics of plasticity arise from the coordinated timing of these interacting representations, as they persist and evolve over time through eligibility traces and network dynamics. Plasticity rules are determined both by the anatomical availability of signals and their temporal orchestration within the circuit. In light of recent advances in connectomics, this workshop will explore how network architecture functions as a structural prior over learning dynamics, biasing, and constraining synaptic plasticity.
Schedule: Monday, Sept 28
| 14:00 | Emmanouil Giannakakis & Anna-Maria Jürgensen (Imperial College London & University of Cambridge, UK) |
Introduction |
| 14:15 | Wulfram Gerstner (École polytechnique fédérale de Lausanne, Switzerland) |
Beyond Hebb: Self-supervised representation learning with local predictive plasticity rule |
| 15:00 | Loreen Hertäg (Technische Universität Berlin, Germany) |
Shaping prediction-error circuits: the role of inhibitory plasticity and connectivity |
| 16:00 | Coffee break | |
| 16:30 | Claire Meissner-Bernard (Sorbonne University, France) |
Formation and properties of memory networks with excitatory-inhibitory assemblies |
| 17:15 | Tim Vogels (Institute of Science and Technology Austria (ISTA), Austria) |
TBA |
| 18:00 | Emmanouil Giannakakis & Anna-Maria Jürgensen (Imperial College London & University of Cambridge, UK) |
Concluding remarks and discussion |
Schedule: Tuesday, Sept 29
| 8:30 | Emmanouil Giannakakis & Anna-Maria Jürgensen (Imperial College London & University of Cambridge, UK) |
Introduction |
| 8:35 | André Fiala (University of Göttingen, Germany) |
Beyond simple associations: Dissecting neuronal circuits and dynamics of higher-order learning in Drosophila |
| 9:15 | Albert Cardona (University of Cambridge & MRC Laboratory of Molecular Medicine Cambridge, UK) |
The role of axo-axonic synapses in signal processing |
| 10:00 | Coffee break | |
| 10:30 | Panayiota Poirazi (Foundation for Research and Technology Hellas, Greece) |
Exploring the role of synaptic and dendritic plasticity in flexible learning |
| 11:15 | Andreas Lüthi (Friedrich Miescher Institute for Biomedical Research (FMI), Switzerland) |
Dynamic interactions between memory traces during higher order learning |
| 12:00 | Emmanouil Giannakakis, Anna-Maria Jürgensen (Imperial College London & University of Cambridge, UK) |
Concluding remarks and discussion |
Advances in optimization of biologically constrained models and how to use them
Organizers:
Alessio Quaresima | Institut de l’Audition – Institut Pasteur, France
Julia Gygax | Friedrich Miescher Institute for Biomedical Research (FMI), Switzerland
Maciej Kania | Institute of Science and Technology Austria (ISTA), Austria
Zoe Harrington | Institute of Science and Technology Austria (ISTA), Austria
Description
Models in computational neuroscience provide understanding at different levels. Abstract models with numerically tractable algorithms can be optimized to reproduce experimental data, but they often lack direct interpretability of the underlying biological mechanisms. Conversely, biologically constrained models provide mechanistic insights into brain circuits. Still, it is challenging to achieve computations or reproduce data with them, as they cannot be explicitly trained due to their non-differentiable dynamics or overall complexity. In this workshop, we will present novel approaches to model optimization that combine the best of both worlds to gain insights into brain computations. The methods to optimize biologically constrained models often fall into two groups. On the one hand, the model parameter space is explored algorithmically using high-throughput approaches and tested against experimental data. Unlike traditional “hand-tuned” parameter exploration, these novel methods offer entire families of solutions. Furthermore, they identify degeneracies in the parameter space that can lead to testable experimental predictions. On the other hand, one can optimize models to solve a task. Adding biological constraints steers them to exhibit realistic activity patterns while performing a specific function. This approach enables us to identify the role of different biological constraints on computations and how they shape the model’s solution to the task. Given the rapid advances in optimizing biologically constrained models, it is necessary to provide a comprehensive overview of the leading methods, their respective advantages and limitations, and the research questions for which they are best suited. This workshop will feature a series of talks ranging from gradient-free optimization to surrogate gradients and Bayesian inference. Our goal is to provide conceptual frameworks and tools for investigating complex systems with no obvious analytical solutions. We will discuss how the identified solution space should be interpreted in the context of the optimization method used and how this advances our understanding of brain computations.
Schedule: Monday, Sept 28
| 14:00 | Maciej Kania (Institute of Science and Technology Austria (ISTA), Austria) |
Introduction |
| 14:05 | Julia Gygax (Friedrich Miescher Institute for Biomedical Research (FMI), Switzerland) |
Theory behind surrogate gradients and how to use them to study neuronal assemblies |
| 14:35 | Guillaume Bellec (TU Wien, Austria) |
Perturbation testing to validate deep learning models of cortical computation |
| 15:05 | Rory Bedford (Friedrich Miescher Institute for Biomedical Research (FMI), Switzerland) |
Connectome-constrained spiking network models of functional activity |
| 15:35 | Dan Goodman (Imperial College London, UK) |
Learning to use neuromodulation for efficient sensory processing |
| 16:00 | Coffee break | |
| 16:30 | Robert Gütig (Charité – Universitätsmedizin Berlin, Germany) |
Interactions between long- and short-term synaptic plasticity transform temporal neural representations into spatial |
| 17:00 | Jonathan Cornford (University of Leeds, UK) |
Mirror descent as a framework for normative brain-like learning |
| 17:30 | Julijana Gjorgjieva (Technical University of Munich, Germany) |
Discovering biologically plausible rules from trained RNNs |
| 18:00 | Discussion | |
Schedule: Tuesday, Sept 29
| 8:30 | Maciej Kania (Institute of Science and Technology Austria (ISTA), Austria) |
Introduction |
| 8:35 | Jakob Macke (Max Planck Institute for Intelligent Systems, University of Tübingen & Tübingen AI Center, Germany) |
Simulation-based inference: Progress, promise, open problems |
| 9:05 | Zoe Harrington (Institute of Science and Technology Austria (ISTA), Austria) |
Geometric and topological analysis of SBI posteriors over plasticity rule space in spiking networks |
| 9:35 | Anastasia Krouglova (VIB-Neuroelectronics Research Flanders (NERF) & KU Leuven, Belgium) |
Multifidelity simulation-based inference for computationally expensive simulators |
| 10:00 | Coffee break | |
| 10:30 | Alessio Quaresima (Institut de l’Audition – Institut Pasteur, France) |
Data-optimized biophysical model identifies parallel functional subnetworks |
| 11:00 | Helmut Strey (State University of New York at Stony Brook & Massachusetts Institute of Technology, USA) |
Scientific machine learning of chaotic systems discovers governing equations for neural populations |
| 11:30 | Adrienne Fairhall (University of Washington, USA) |
Meta-learning learning rules that build and maintain circuit motifs |
| 12:00 | Discussion | |
Bridging the species divide: comparative approaches to study neural signatures
Organizers:
Natalie Schaworonkow | Ernst Strüngmann Institute, Germany
Dennis Nestvogel | Max Planck Institute of Psychiatry, Germany
Description
Neuroscience has become immensely data-rich, with high-density recordings across many species. But animal and human research remains largely siloed, even within the same model system. Cross-species comparison of electrophysiological activity offers a promising path to establish necessary and sufficient conditions for linking circuit dynamics and behavior, leveraging the excellent temporal resolution and access to mesoscale dynamics that electrophysiology provides. Many neural signatures appear similar across species, but what does it mean to equate rhythms or transient events between different species? This challenge is compounded by the fact that even within the same species and brain region, electrophysiological signatures are often labeled inconsistently across studies. For example, the alpha rhythm in the mouse has been proposed to occur at a frequency of 3–5 Hz, which differs significantly from the canonical range of 8–13 Hz observed in humans. Furthermore, there are examples of abrupt relabeling, such as the reclassification of posterior 8 Hz activity from theta to alpha during human development. In this workshop, we examine electrophysiological signatures across species through specific examples (theta, alpha, beta, and gamma activity). We ask whether we can establish common definitions and terminology based on criteria beyond frequency alone, including spatial distribution, generative mechanisms at the cellular level, and functional modulation by behavior. We hope to stimulate discussion about the minimal conditions necessary to identify functionally equivalent electrophysiological signatures across species and, through these signatures, investigate behavior and cognition.
Schedule: Monday, Sept 28
| 14:00 | Natalie Schaworonkow (Ernst Strüngmann Institute, Germany) |
General Introduction |
| 14:30 | Lukas Kunz (University Hospital Bonn, Germany) |
Investigating the relationship between theta and spiking activity through human single-unit recordings |
| 15:15 | Abraham Z. Vollan (Kavli Institute for Systems Neuroscience, Norway) |
Adaptive modulation of theta sweeps in the rodent spatial navigation circuit |
| 16:00 | Coffee break | |
| 16:30 | Tzvetan Popov (University of Zurich, Switzerland & University of Konstanz, Germany) |
Alpha rhythms in honey bees and humans |
| 17:10 | Alina Studenova (Max Planck Institute for Human Cognitive and Brain Sciences, Germany) |
On the relevance of alpha oscillations in generation of evoked responses |
| 17:50 | Sacha van Albada (University of Cologne & Forschungszentrum Jülich, Germany) |
Mean-field and spiking models of thalamocortical alpha rhythm generation |
Schedule: Tuesday, Sept 29
| 8:30 | Recap | |
| 8:40 | Dennis Nestvogel (Max Planck Institute of Psychiatry, Germany) |
Dissecting thalamocortical dynamics of an alpha-like rhythm in the mouse |
| 9:20 | Julia Veit (University of Bremen & University of Freiburg, Germany) |
Inhibitory control of cortical gamma synchronization |
| 10:00 | Coffee break | |
| 10:30 | Eric Drebitz (University of Bremen, Germany) |
Gamma synchronization between neurons in the visual cortex is causal for effective information processing and behavior |
| 11:10 | Anton Sirota (LMU Munich, Germany) |
Anatomical and spectral dissection of high frequency oscillations – towards detection and sorting of units of circuit computation. |
| 11:50 | General discussion | |
From lab to wild: how internal states enable adaptive behavior
Organizers:
Claire Sturgill | TUD Dresden University of Technology, Germany
Arman Behrad | TUD Dresden University of Technology, Germany
Description
Brains are not purely stimulus–response machines; instead, behavior is profoundly shaped by internal states—hidden, temporally persistent variables that modulate how sensory inputs are processed and transformed into behavioral output. These hidden processes include, for instance, arousal, motivation, attention, affect, and homeostatic drives, and integrate information from the external environment and physiological conditions into centralized brain states that orchestrate behavior and physiology (Marques, …, Li 2019). A central challenge is that the same behavioral output can arise from different internal conditions, and conversely, the same sensory input can lead to different actions depending on an organism’s internal state—especially in naturalistic and ecological settings where task epochs and experimenter-defined labels are absent (Gupta et al. 2024). Signatures of such internal processes are often studied with diverse terminologies: “internal states” in systems neuroscience, “internal values or decisions” in ecological approaches to neuroscience, or “internal cognitive processes” in cognitive neuroscience. In this workshop, we will reconcile these parallel perspectives, establishing shared definitions and benchmarks, and fostering collaboration between experimental and theoretical communities working on different facets of internal states in the brain and behavior. In this workshop, we will mainly focus on two approaches to internal state: (1) Naturalistic behavioral and physiological readouts. Internal states are inferred from changes in movement kinematics, engagement, exploration/exploitation structure, and multimodal neurophysiological signals. (2) Neural population dynamics, where internal states manifest as low-dimensional manifolds, attractors, or transitions between input-driven and autonomous regimes. By bringing these perspectives together, we will aim to define common principles, evaluation criteria, and open challenges for identifying internal states across neural systems.
References:
Marques, J. C., Li, M., Schaak, D., Robson, D. N., & Li, J. M. (2020). Internal state dynamics shape brainwide activity and foraging behaviour. Nature, 577(7789), 239-243.
Gupta, D., DePasquale, B., Kopec, C. D., & Brody, C. D. (2024). Trial-history biases in evidence accumulation can give rise to apparent lapses in decision-making. Nature communications, 15(1), 662.
Schedule: Monday, Sept 28
| 14:00 | Jonathan Pillow (Princeton University, USA) |
Characterizing the Emergence of Internal States across Learning |
| 14:40 | Jennifer Mengbo Li (Max Planck Institute for Biological Cybernetics, Germany) |
Brainwide joint analysis of neuromodulatory and cognitive networks during spatial learning in freely swimming zebrafish |
| 15:20 | Shervin Safavi (TUD Dresden University of Technology, Germany) |
Neural, computational, and behavioral mechanisms of internal decision processes |
| 16:00 | Coffee break | |
| 16:30 | Diksha Gupta (University College London, UK) |
Redundancy and robustness in decision-making circuits |
| 17:10 | Saurabh Vyas (Carnegie Mellon University, USA) |
Flexible problem solving in prefrontal cortex |
| 17:50 | Jessica Cardin (Yale University, USA) |
Multimodal models of cortical dynamics: Neuromodulatory signaling and internal state |
Schedule: Tuesday, Sept 29
| 8:30 | Thomas Luo (University of Utah, USA) |
A neural marker of internal decision commitment |
| 9:10 | Cindy Poo (Allen Institute for Neural Dynamics, USA) |
Adaptive behavior in a dynamic patch foraging environment |
| 10:00 | Coffee break | |
| 10:30 | Tim Buschman (Princeton University, USA) |
Geometry of cognitive flexibility |
| 11:10 | Roxana Zeraati (Max Planck Institute for Biological Cybernetics, Germany) |
Optimal foraging under naturalistic temporal dynamics |
| 11:50 | Speaker panel | |
Reconciling biology and function in large-scale brain models
Organizers:
Sacha van Albada | Forschungszentrum Jülich & University of Cologne, Germany
Hans Ekkehard Plesser | Norwegian University of Life Sciences, Norway
Description
Integrating biological constraints with realistic functional properties in neural network models—from learning to behavioral output—is a long-standing challenge in computational neuroscience. “Top-down” models of learning and function generally incorporate some biologically plausible aspects, but usually not to an extent that allows for testing these models directly against biological data or for disambiguating between potential underlying biological mechanisms. With “bottom-up” models systematically derived from biological data, on the other hand, scientists rarely investigate which network functions the models may support. However, recent years have brought considerable progress in the development of brain models that capture the cytoarchitecture, connectivity, and electrophysiology of large portions of the mammalian brain at cellular resolution. While the large-scale nature of these models is not a unique requirement for their biological plausibility, small models tend to focus on a few select phenomena besides distorting aspects of dynamics and learning, highlighting a need to move beyond minimal models. Integrative large-scale models can, in principle, serve as increasingly reliable test beds for theories of learning and information processing in the brain. The problem is then how we can reconcile the bottom-up and top-down approaches to arrive at an integrated understanding of brain function. In this workshop, we will address how to integrate structure, dynamics, and function in network models by considering the following questions: What are the scientific, technical, and organizational hurdles in bringing together biological constraints with realistic network function? What are some promising approaches? How can we tackle the problem in a way that not only reproduces available data but also generates reliable predictions? How can we arrive at well-constrained models suitable for disambiguating between different theories of brain function? How can we adapt learning rules proposed in more artificial settings (e.g., non-spiking models, feedforward networks, small-scale networks) to biological constraints? How can we develop integrative models that combine various functional properties? Guided by these questions, we hope to take a step forward toward a practical work program for reconciling biology and function in large-scale brain models.
Schedule: Monday, Sept 28
| 14:00 | Jun Igarashi (RIKEN, Japan) |
Oscillatory activity in large-scale simulations of connectome-constrained spiking neural network models of the cerebral cortex |
| 14:30 | Lies Van Dael (Forschungszentrum Jülich, Germany) |
Multi-area spiking network model of macaque cortex with joint excitatory/inhibitory clusters |
| 15:00 | Ján Antolík (Charles University, Czech Republic) |
Bridging spontaneous, visually evoked, and artificially evoked states with a single underlying substrate |
| 15:30 | Discussion | |
| 16:00 | Coffee break | |
| 16:30 | Friedemann Zenke (Friedrich Miescher Institute & University of Basel) |
Closing the loop: from representation learning to local plasticity and back |
| 17:00 | Caroline Haimerl (Champalimaud Foundation, Portugal) |
From action to abstraction: multiscale representation learning for perception and action |
| 17:30 | Discussion | |
Schedule: Tuesday, Sept 29
| 8:30 | Jakob Macke (University of Tübingen, Germany) |
Learning mechanistic models from neurons to networks to computations |
| 9:00 | Mihai A. Petrovici (University of Bern, Switzerland) |
Error transport in cortical microcircuits |
| 9:30 | Mario Senden (Maastricht University, The Netherlands) |
Reconciling biology and function via biophysics-informed deep learning |
| 10:00 | Coffee break | |
| 10:30 | Laureline Logiaco (University of Colorado Anschutz, USA) |
A multi-region neural network with complementary model-free and model-based learning systems unifies sensorimotor phenomena |
| 11:00 | Rick Granger (Dartmouth College, USA) |
From physiological computation to mechanistic neuropharmacology |
| 11:30 | Markus Diesmann (Forschungszentrum Jülich, Germany) |
Building on models: past, present, perspectives |
| 12:00 | Discussion | |
Quo vadis, neural network theory?
Organizers:
Alexander van Meegen | RWTH Aachen University, Germany
Jacob Zavatone-Veth | Harvard University, USA
Description
Driven by the remarkable progress of artificial intelligence, the theory of artificial neural networks has advanced substantially in recent years. These developments promise a deeper understanding of neural networks more broadly, including biological neural networks. Theoretical insights may inspire novel analysis methods to address the growing deluge of data, helping us make sense of large-scale recordings and reconstructions. Furthermore, they may open new perspectives and questions, motivating new experiments as well as reinterpretations of existing data. The guiding question of this workshop is how theoretical advances for artificial networks can be translated into the greatest impact on neuroscience. As a first step, the current status quo will be assessed through invited speakers who are at the forefront of neural network theory. All speakers will conclude their talks with a view toward the future: how do they envision the impact of their work? These visions will then be critically discussed with the audience, considering both the promises and the potential pitfalls. As a synthesis, the workshop will conclude with a panel discussion focused on the broader perspective.
Schedule: Monday, Sept 28
| 14:00 | Sebastian Goldt (Scuola Internazionale Superiore di Studi Avanzati (SISSA), Italy) |
Diverse perceptual biases emerge from Hebbian plasticity in a recurrent neural network model |
| 14:40 | Valentin Schmutz (University of Oxford, UK) |
High-dimensional neuronal activity from low-dimensional latent dynamics: a solvable model |
| 15:20 | Friedrich Schüßler (TU Berlin, Germany) |
A theory of initialisation’s impact on specialisation |
| 16:00 | Coffee break | |
| 16:30 | Agostina Palmigiano (University College London, UK) |
A unified theory of feature learning in RNNs and DNNs |
| 17:10 | Lorenzo Tiberi (Harvard University, USA) |
Manifold geometry underlies a unified code for category and category-independent features |
| 17:50 | SueYeon Chung (Harvard University, USA) |
Neural population geometry and optimal coding of tasks with shared latent structure |
Schedule: Tuesday, Sept 29
| 8:30 | Tatiana Engel (Princeton University, USA) |
The dynamics and geometry of choice in the premotor cortex |
| 9:10 | David Clark (Harvard University, USA) |
Dynamical Theory of Optimized Structures in Large Recurrent Neural Networks |
| 10:00 | Coffee break | |
| 10:30 | Francesca Mignacco (Princeton University, USA) |
Optimal protocols for continual learning via statistical physics and control theory |
| 11:10 | Kanaka Rajan (Harvard University, USA) |
Measuring and controlling solution degeneracy across task-trained recurrent neural networks |
| 11:50 | Panel discussion | |
Going wild – Cognition, behaviour, and neural processing in natural and naturalistic settings
Organizers:
Fred Wolf | Max Planck Institute for Dynamics and Self-Organization & University of Göttingen, Germany
Jan Benda | University of Tübingen, Germany
Kerstin Schmidt | Federal University of Rio Grande do Norte (UFRN Natal, Brazil)
Description
Animal and human brains evolved to process information and organize behaviour in complex dynamic ecosystems such as primary forests, primal river basins, and similar habitats set in a world that was largely not yet shaped by near-universal anthropogenic change. Over the past decade, neurotechnology, optical instrumentation, and revolutionary advances in information technology have opened exciting avenues to observe, interact with, and quantitatively measure behavior, physiology, and neural function in unrestrained animals in complex environments. Perhaps the most revolutionary current frontier in this research direction is driven by studies of cognition, behaviour, and neural function in maximally naturalistic settings and wild natural habitats. Increasingly, a new generation of study designs integrates field-based and lab-based approaches to uncover the mechanisms and dynamics of exactly the behaviours that animal brains evolved to perform best. This emerging field of “natural neuroscience” studies raises novel and unique challenges in a wide range of domains, from big data methods to the development of mobile and miniature devices and to computational modelling and fundamental theory. The workshop “Going Wild – Cognition, Behaviour and Neural Processing in Natural and Naturalistic Settings” is designed to provide a platform for projects at the forefront of this emerging field and to bring together experts rooted in behavioural ecology and evolution with computational and systems neuroscientists. Topics covered range from primate cognition, navigation, and social foraging to prey capture and the coordination of animal group behaviour. The studies presented, as well as a dedicated discussion session, will offer ample opportunities to learn about and exchange on the novel computational, systems, and data-science challenges arising with this emergent research direction. On an equal footing, we will facilitate discussions on fundamental questions such as “What really defines an ‘ecologically valid’ behavioural task in lab and field research?” and “Are there novel ethical dimensions for responsible animal research arising in studies conducted in wild and natural settings?”.
Schedule: Monday, Sept 28
Session: Mammals
| 14:00 | Jan Benda & Fred Wolf (University of Tübingen / Max Planck Institute for Dynamics and Self-Organization & University of Göttingen, Germany) |
Introduction |
| 14:10 | Daniel Takahashi (Federal University of Rio Grande do Norte (UFRN Natal), Brazil) |
Multiscale embedding of neural signal for generative analysis of natural behavior |
| 14:50 | Irene Lacal (German Primate Center – Leibniz Institute for Primate Research (DPZ), Germany) |
Deciding while walking – the planning horizon of primate premotor cortex |
| 15:25 | Cory Miller (University of California, San Diego, USA) |
Active vision in marmoset arboreal prey capture |
| 16:00 | Coffee break | |
| 16:30 | Kerstin Schmidt (Federal University of Rio Grande do Norte (UFRN Natal), Brazil) |
Adapting the visual cortex to the natural world: Organization and diversity |
| 17:00 | Ilia Leonov (Max Planck Institute for Dynamics and Self-Organization & University of Göttingen, Germany) |
Decoding the dynamics of free goal-directed behavior |
| 17:30 | Ahmed El Hady (Max Planck Institute for Animal Behavior, Germany) |
Towards a general mechanistic theory of foraging |
| 18:00 | Ali Nourizonoz (University of Geneva, Switzerland) |
Optimized planning of a high-risk behavior in a miniature nocturnal primate |
Schedule: Tuesday, Sept 29
| 8:30 | Jason Kerr (Max Planck Institute for Neurobiology of Behavior – caesar, Germany) |
Vision for action in hunting Harris hawks |
| 9:15 | Jerome Baron (Federal University of Minas Gerais, Brazil) |
Rethinking parallel visual pathways through the eyes of ‘nocturnal’ birds |
| 10:00 | Coffee break | |
| 10:30 | David de Santana (Museu Paraense Emílio Goeldi, Brazil) |
Adaptive radiation without anatomical radiation in electric eels |
| 11:10 | Lena Stöckl & Jan Benda (University of Tübingen, Germany) |
Musical electric fish – from neotropical rivers into the lab |
| 11:50 | General discussion | |
For further questions, please contact us at: bernstein.conference@fz-juelich.de.
