Bernstein Conference 2025

Location

Frankfurt am Main

Date

Sep 29 – Oct 2

Abstracts

Repository

Photographic retrospective

Previous Next

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

Invited Talks

Alison Barker | Max Planck Institute for Brain Research, Germany
Dynamics of social systems: Communication and cooperation in the naked mole-rat (abstract)

Upinder S. Bhalla | National Centre for Biological Sciences, India
Mismatch detection through molecules-to-network computations (abstract)

Albert Compte | Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Spain
Prefrontal attractor dynamics of dual task learning (abstract)

Luca Mazzucato | University of Oregon, USA
Control of population activity in prefrontal cortex via microstimulations (abstract)

Tirin Moore | Stanford University, USA
Short-term coding of remembered stimuli: Lessons from large-scale electrophysiology in the primate brain (abstract)

Edvard Moser | Norwegian University of Science and Technology, Norway
Network coding in grid cells and place cells: From space to memory (abstract)

Katherine Nagel | NYU School of Medicine, USA
A circuit and synaptic architecture for working memory during olfactory navigation (abstract)

Panayiota Poirazi | Foundation of Research and Technology-Hellas (FORTH), Greece
How dendrites empower learning in biological and artificial brains (abstract)

Viola Priesemann | Max Planck Institute for Dynamics and Self-Organization, Germany
Dendritic balance, synaptic design and emergence of information processing (abstract)

Andreas Tolias | Stanford University, USA
Foundation models and digital twins of the brain (abstract)

Valentin Braitenberg Award Winner

Sara A. Solla | Northwestern University, USA

Contributed Talks

Denis Alevi | Yale University, USA
Representational drift as a correlate of memory consolidation (abstract)

Deyue Kong | Frankfurt Institute for Advanced Studies, Germany
Single-cell optogenetic perturbations reveal stimulus-dependent network interactions in ferret V1 (abstract)

Daniel Levenstein | Yale University, USA
Sequential predictive learning is a unifying theory for hippocampal representation and replay (abstract)

Jure Majnik | Aix-Marseille University / Inserm / INMED / Turing Center for Living Systems, France
Tracking and perturbing developmental trajectories in postnatal mouse neocortex (abstract)

Haleigh Mulholland | University of Minnesota, USA
Large-scale recurrent networks selectively stabilize neural activity in visual cortex (abstract)

Lorenzo Posani | Columbia University, USA
Rarely categorical, always high-dimensional: How the neural code changes along the cortical hierarchy (abstract)

Lynn K. A. Sörensen | Massachusetts Institute of Technology, USA
Hierarchical optimization predicts plasticity in the macaque inferior temporal cortex following object training (abstract)

Ben von Hünerbein | University of Bern, Switzerland
ELiSe: Efficient Learning of Sequences in Structured Recurrent Networks (abstract)

Satellite Workshops

Emerging networks: Computational approaches to brain development
Organizers: Julijana Gjiorgjieva, Ruben A. Tikidji-Hamburyan

Description:

In computational and theoretical neuroscience, brain networks are often treated as fixed structures—be it dendritic morphology, neuronal excitability, axonal projections, or whole-brain connectomes. Yet, all hallmark features of the adult brain—such as sensory topographies, orientation selectivity in the visual cortex, and grid cells in the entorhinal cortex—emerge and are shaped during early developmental stages. These refinement processes are orchestrated by a combination of genetic programming and activity-dependent mechanisms. Understanding this process is of key importance not just for understanding the normal functioning of the brain, but also for unraveling cellular and network disruptions that lead to neurodevelopmental disorders.

This workshop focuses on the dynamic processes that underlie the formation and refinement of brain circuits. Development is not merely a transitional phase toward the mature brain; rather, developing networks possess unique properties tailored to support early neural activity, guide circuit formation, and suppress maladaptive patterns. For instance, many neural circuits generate specific patterns of spontaneous activity even before the onset of sensory experience, which may serve as training data to shape connectivity and prepare these networks for adult function. To understand these processes, researchers employ a wide range of computational models—from abstract frameworks to detailed biophysical simulations. This growing field of developmental systems neuroscience seeks to uncover how structure and function co-emerge during development, and how early experience influences the trajectory of brain maturation. This workshop will cover topics and include models of synaptic and dendritic development, the generation and propagation of spontaneous activity, activity-driven refinement of connectivity, and early sensory and cognitive circuit formation. From dendritic organization of synapses on pyramidal neurons to signal propagation in immature thalamic circuits, this workshop offers a comprehensive overview of the latest advances and future directions in modeling the developing brain.

Confirmed speakers:

Neural architectures in action: Linking structure to computation across scales
Organizers: Pascal Nieters, Johanna Senk, Susanne Kunkel

Description:

With increasingly detailed data on neural circuitry structures becoming available, the computational significance of these intricate architectures has become central to neuroscience research. This workshop explores computational models and theories across multiple scales—from dendritic computation in single neurons to microcircuits and brain-scale networks—that leverage these architectural insights to develop novel computational paradigms, address longstanding questions, and propose new explanations of brain function. Speakers will highlight how structural motifs may represent fundamental computational primitives for both theoretical understanding and practical applications. Special emphasis will be placed on how structural plasticity mechanisms adjust and adapt these structures to enable computation and learning beyond traditional synaptic weight adjustments, on the use of dendritic compartmentalization in learning and computation, and on network motifs in lateral and large-scale networks. We bring together researchers across neuroscience disciplines to advance our understanding of how neural architecture shapes computation in biological, artificial, and neuromorphic systems.

Confirmed speakers:

Toward a joint definition of neural-behavioral states
Organizers: Arman Behrad, Shervin Safavi

Description:

Understanding brain states requires more than the isolated study of either neural dynamics or behavior — it demands an integrative framework that acknowledges their complex, reciprocal interaction. For instance, this perspective led to developing key methods in the field (e.g., CEBRA, Schneider et al, Nature 2023) of shared latent states that underlie observable dynamics and behavioral outputs, offering a more comprehensive and biologically grounded definition of brain states. In this workshop, we bring together computational and experimental neuroscientists studying brain states with such an integrative perspective, and we hope that through this exchange, we find a clearer idea of a joint definition of brain-behavior state. We aim to stimulate discussions on the benefits and challenges of different frameworks and perspectives, and how they can facilitate a multi-faceted understanding of brain computation (consequently, behavior) in relation to neural dynamics. We believe that through this exchange of ideas, we can find new paths toward systemizing an approach to define brain state jointly based on neural dynamics and behavior (or at least the pros and cons of each approach). Moreover, the complementary perspectives of our speakers, synergized with discussions in our workshop, would promote new research directions and collaborations that will broaden the extent of brain research to gain a more holistic understanding of brain computation, dynamics, and behavior. Our speakers will discuss diverse perspectives to jointly characterize states for information processing systems (both biological and artificial). These range from well-established (yet diverse) approaches such as task-optimized RNNs (Valerio Mante), neurophysiology (Dennis Nestvogel), and learning low-dimensional embedding from joint neural and behavioral data (Xiaohang Yu’s CEBRA 2.0 from Mackenzie Mathis lab; Maryam Shanechi’s work, e.g., DPAD published in Nat Neuro 2024, and BRAID published in ICLR 2025), to less explored approaches that put more weight on behavior (recent work of Tatiana Engel and Meming Park). Our speakers also cover novel approaches that exploit recent developments in data-driven dynamical systems to introduce a common language to study states both in biological and artificial systems (Arman Behrad from Shervin Safavi’s lab). Moreover, our workshop benefits from recent developments in experimental neuroscience to have new perspectives on the joint definition of neural-behavioral state. First, through the presentation of Virginia Ruetten (from Misha Ahrens’s lab in Janelia campus), who intends to define the joint state, by recording neurons simultaneously with body ‘cells’ (which is a novel and unique approach), and to empirically find such a joint state in their data.

Confirmed speakers:

Same words, different worlds: Conceptual consistency in systems neuroscience
Organizers: Mattia Chini, Irina Pochinok, Simon Musall

Description:

This workshop brings together philosophers, experimentalists, theorists, and engineers to tackle an important but often overlooked source of variability across studies in systems neuroscience: conceptual and methodological drift. While there are growing efforts to establish standardized preprocessing or statistical modeling approaches across labs, much less attention is paid to the impact of inconsistent use of core concepts when researchers use the same words to mean slightly different things. Concepts like “causality”, “connectivity”, or “representation” are central to how we formulate questions, choose experimental and analytical tools, and interpret results. Yet these terms often carry different meanings across subfields, especially in an inherently interdisciplinary field, such as systems neuroscience. This heterogeneity carries a high risk of conceptual fragmentation, leading to distinct analytical strategies and ultimately to incompatible results. To promote scientific integrity and reproducibility, it is therefore crucial to identify and understand the key conceptual differences that shape how scientific conclusions are drawn. The workshop directly tackles this challenge in its full complexity. Speakers will examine how abstract concepts such as causality or modularity are conceptually defined, and how these definitions can be operationalized. The workshop will also cover seemingly concrete phenomena, such as oscillations, sharp wave-ripples, and intrinsic timescales, to examine how even well-established phenomena are shaped by conceptual assumptions and methodological decisions. Talks will range from theory to experimental practice, including the logic behind analysis standardization tools and the epistemology of mechanistic explanations. The format and topic of the workshop are well-suited to promote active participation and open debate. By design, the workshop aims to surface controversial issues and encourage discussing competing approaches and implicit assumptions.

Confirmed speakers:

Representational drift and its consequences for learning and memory
Organizers: Jens-Bastian Eppler, Alex Roxin

Description:

Recent years have revealed that neuronal representations are not stable, but instead change over time, a phenomenon known as representational drift. This workshop aims to explore the consequences of this drift for learning, memory consolidation, and behavioral stability. We will bring together experimental and theoretical perspectives to address key questions: How does representational drift manifest across brain areas and tasks? How can the brain maintain stable behavior and memory performance despite ongoing changes in neuronal activity? And what mechanisms might underlie the coexistence of plasticity and stability across different levels of brain organization? The workshop will feature eight talks by invited speakers, complemented by an introductory overview and a concluding summary provided by the organizers, as well as a dedicated discussion session to foster interaction and synthesis across disciplines.

Confirmed speakers:

Machine learning advances for constraining interpretable models of dynamics from brain recordings
Organizers: Richard Gao, Manuel Brenner

Description:

The goal of the workshop is to bring together computational modelers and machine learning (ML) researchers to explore how ML methods can accelerate the development of interpretable dynamical models constrained by brain activity recordings at multiple scales. Emphasizing both mechanistic models (e.g., Hodgkin-Huxley, IF spiking neuron networks, neural mass models) and structured ML approaches (e.g., SLDS, piecewise linear RNNs, LFADS), the workshop aims to bridge the gap between biophysical realism and data-driven modeling. Key considerations for relevant contributions include: (1) models must be explicitly constrained to reproduce recorded brain dynamics, going beyond purely task-trained models, and (2) models must be interpretable—either through mechanistic design (e.g., structured connectivity, biophysical constraints, piecewise linearity) or through post hoc analysis of learned dynamics. The workshop will survey different inference strategies, ranging from automatic differentiation and evolutionary algorithms to classical and simulation-based Bayesian inference, that enable efficient and biologically grounded model fitting. Through presentations and discussions, we hope to address common challenges in model design, inference, analysis, and evaluation, identifying paths forward for integrating ML advances with interpretable brain dynamics modeling.

Confirmed speakers:

From molecules to networks: The dendritic processes that shape learning and memory
Organizers: Björn Kampa, Gaia Tavosanis, Willem Wybo

Description:

Neurons collect their synaptic input on large dendritic trees wherein these inputs are also integrated and processed. These mechanisms have been described already for several decades and brought forward many promising computational advantages. Yet, it has been the recent advance in optophysiological techniques that enables to measure dendritic activity in the intact brain of animals performing complex behavior tasks. New activity sensors, genetically engineered and tailored to report local dendritic signals opened new vistas and offer promising new research directions. At the same time, incorporating dendrites into deep neural networks has shown to improve their performance, reduce their energy requirements and allow learning continuously new content without forgetting previously acquired knowledge. The workshop brings together leading young and senior experts from the field covering experimental and computational work from the molecular to the systems level leading to new fundamental models and AI networks. The aim of the workshop is to exchange recently gained knowledge in a series of talks and to create new ideas in intense discussions.

Confirmed speakers:

Control theory approaches for analysing, modeling, and manipulating brain activity and cognitive function
Organizers: Luca Mazzucato, Dmitri Chklovskii, Samir Suweis

Description:

Current neurotechnology enables artificial perturbations of single neurons in awake animals. Advancing this promising avenue to achieve targeted manipulation of brain activity will be a stepping stone in developing new brain-computer interfaces (BCIs) to ameliorate cognitive dysfunctions, such as in deep brain stimulation for Parkinson’s disease, post-traumatic stress disorder, obsessive-compulsive disorder, treatment-resistant depression, and addiction. However, most procedures apply open-loop stimulations blindly to the specific properties of the stimulated neurons. A major theoretical obstacle is that, although artificial perturbations can affect behavior and task performance, it is not known how to predict and model their effects, as current approaches are based on inefficient trial-and-error procedures. We identify control theory as a promising approach to address these pressing challenges and bridge from theoretical modeling to neural engineering to translational applications. At the same time, we believe that control theory could provide a principled approach to analyze and understand how information is processed and transmitted in the brain to enable perception, cognition, and action. Control theory could provide a novel and principled approach to understanding the precise input/output relationships of cortical circuit dynamics and function from the lens of perturbations or multi-area interaction. In this workshop, we will bring together leading experts in control theory and perturbation approaches to brain circuits at different levels of description, from non-invasive whole-brain dynamics in humans to invasive approaches in animal models. Our goal is to identify outstanding challenges and share methods from different backgrounds to chart a path forward for the future applications of brain controllability, bringing together perspectives from dynamical systems theory, control theory, and information theory.

Confirmed speakers:

Top-down control of neural dynamics
Organizers: Aitor Morales-Gregorio, Anno Christopher Kurth

Description:

The mammalian cerebral cortex can be subdivided into multiple areas on both anatomical and functional grounds. These areas are organized into multiple parallel pathways integrating information along various streams, which ultimately underlie brain function as a whole. Top-down control of areas from regions that are located higher in the cortical hierarchy can have a profound impact on cortical dynamics. This may, for example, take the form of altering network states via selective activation of sub-networks, the expansion of dimensionality of neural activity, the modulation of ON-OFF dynamics by selective attention, or the modulation of timescales, facilitating context-dependent information processing. In the workshop, we want to bring together findings from experimental studies and computational modeling to discuss the role of as well as the mechanisms behind top-down control of neural dynamics.

Confirmed speakers:

Relational Inference and knowledge composition via neuronal geometric representations
Organizers: Sofia Raglio, Maurizio Mattia, Gabriele Di Antonio

Description:

This workshop aims to explore how the brain constructs internal maps of implicit relationships between stimuli, and how task structure shapes the geometry of these representations. At the intersection of cognitive and computational neuroscience, we will discuss key questions about generalisation: What principles allow the brain to infer novel relations from prior knowledge? How are spatial representations intertwined with higher cognition? And can geometry serve as a fundamental coding paradigm for structured information? To address these themes, we bring together researchers approaching the problem from complementary perspectives: computational models of relational generalisation, experimental studies on inferential reasoning, and analyses of the geometric properties of neural representations. Through a series of talks and two panel discussions, we aim to foster dialogue between these communities, identify unifying principles, and outline open challenges that can shape future investigations into the neural basis of abstract, structured cognition.

Confirmed speakers:

Neuromorphic Tug-Of-War v2.0: Neuroscience and AI at different timescales
Organizers: Matteo Saponati, Laura Kriener, Melika Payvand, Filippo Moro, Sebastian Billaudelle

Description:

Neuroscience has played a foundational role in shaping Artificial Intelligence (AI), particularly by inspiring the development of Neuromorphic hardware. However, recent breakthroughs in AI, driven by Machine Learning (ML), have achieved high performance while straying from biological principles. This shift raises ongoing questions about the importance of biological realism in Neuromorphic Computing. In last year’s edition of this workshop, we initiated a community-wide conversation to examine this tension. Following the workshop, we conducted a broad online survey to gather current perspectives from the Neuromorphic Engineering community. The responses revealed a new kind of “tug-of-war”: between short-term and long-term goals, and between top-down theoretical approaches and bottom-up engineering solutions. This year’s workshop builds directly on those findings and explores their implications for the future of the field. We have invited speakers with diverse backgrounds, including Neuroscience, Machine Learning, and Neuromorphic Engineering, who are actively engaged in addressing these questions. They will present recent Neuroscience discoveries that have not yet been fully integrated into Neuromorphic systems, reflect on important lessons from the development of ML, and critically reassess the foundational goals of the field. We will also present quantitative insights from the community survey to anchor the discussion and foster a dialogue aimed at identifying concrete stepping stones for the future of Neuromorphic Computing, both in the short term and in the long term.

Confirmed speakers:

Bernstein Conference 2025

Photographic retrospective

Invited Talks

Valentin Braitenberg Award Winner

Contributed Talks

Satellite Workshops

Follow us on