Bernstein Conference 2024

Location

Frankfurt am Main

Date

Sep 29 – Oct 2

Abstracts

Repository

Photographic Retrospect

Zurück Weiter

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

Invited Talks

Dmitriy Aronov | Columbia University, USA
Neural code for episodic memories in a food-caching bird

Helen Barron | University of Oxford, UK
Building internal models during periods of rest and sleep

Elizabeth Buffalo | University of Washington, USA
Neural dynamics of memory formation in the primate hippocampus

Mark M Churchland | Columbia University, USA
From spikes to factors: understanding large-scale neural computations

Alex Cayco Gajic | Ecole Normale Supérieure, France
Dimensionality reduction beyond neural subspaces

Jan Drugowitsch | Harvard University, USA
Spatial navigation under uncertainty

Jakob Macke | University of Tübingen, Germany
Building mechanistic models of neural computations with simulation-based machine learning

Mala Murthy | Princeton University, USA
Circuit mechanisms for dynamic social interactions

Memming Park | Champalimaud Foundation, Portugal
Forecasting motor cortex activity with a nonlinear latent dynamical system model

Susanne Schreiber | Humboldt Universität zu Berlin, Germany
Cellular action potential generation: a key player in setting the network state

Xiao-Jing Wang | New York University, USA
Distributed dynamics and cognition in the multiregional neocortex

Contributed Talks

Claudia Cusseddu | Technical University of Munich, Germany
A family of synaptic plasticity rules based on spike times produces a diversity of triplet motifs in recurrent networks

Elizabeth A. de Laittre | University of Chicago, USA
Stable cortical coding for a dexterous reach-to-grasp task across motor cortical laminae

Samuel Eckmann | University of Cambridge, UK
Theta-modulated memory encoding and retrieval in recurrent hippocampal circuits

Charles Fieseler | University of Vienna, Austria
Brain-wide manifold-organized hierarchical encoding of behaviors in C. elegans

Fereshteh Lagzi | University of Washington, USA
Reconciling diverse experimental findings on inhibitory tuning in the mouse visual cortex

Charles Micou | University of Cambridge, UK
Sudden tuning curve jumps in cortical representational drift facilitate stable downstream population readouts

Aitor Morales-Gregorio | Charles University, Czechia
State-dependent population activity, dimensionality and communication in the visual cortex

Motahareh Pourrahimi | McGill University, Canada
Human-like behavior and neural representations emerge in a goal-driven model of overt visual search for natural objects

Satellite Workshops

Algorithms and neural mechanisms of social and collective behavior
Organizers: Armin Bahl, Katrin Vogt

Description:

Neuroscience experiments have been classically performed under highly controlled conditions in isolated animals, largely ignoring that, in the wild, many animals usually socially interact in heterogeneous groups. To understand the algorithmic computations and neural mechanisms of such naturalistic behavior, it is therefore important to include social contexts in our theoretical analyses and experimental pipelines. The emerging research field of the study of the algorithms and neural mechanisms of social and collective behavior has recently made major technological breakthroughs, ranging from sophisticated closed-loop virtual reality systems to novel physiological recording methods. These advances make it now feasible to systematically dissect how collective behavior benefits group decision-making performance, which sensory cues animals use to recognize each other, which algorithms govern collective behavior, and how nervous systems combine, potentially conflicting, sensory input. Our proposed satellite workshop aims to bring together leading researchers who have significantly contributed to addressing these questions. We have selected speakers based on an emphasis on a broad perspective across model systems, ranging from insects to fish, to rodents, and primates. This diversity will allow us to identify general principles across species. We have also finely balanced the level of theory and experiment, so the audience can benefit from extensive discussions on developing and applying computational methods to better understand and predict neural and behavioral data.

Confirmed speakers:

Bridging RNNs and data: Hypothesis-testing of network dynamics against neural recordings
Organizers: Isabel M. Cornacchia, Arthur Pellegrino

Description:

Recurrent neural networks (RNNs) have become an increasingly popular tool for modelling the neural computations underlying behavioural tasks. By varying the task they are trained on, their architecture, or their learning rule, a myriad of different models can be created, each corresponding to specific assumptions regarding the neural circuit solving the task. It is, however, not always clear how to test the hypotheses generated by different RNNs against experimental data. Recently, dimensionality reduction methods have been highlighted as candidate tools to compare the neural activity of network models to large-scale neural recordings, opening a window into bridging artificial and biological network dynamics. The present workshop aims to foster discussion between theorists working on neural network models and experimentalists working on neural population recordings to make a step towards systematic hypothesis-testing of the neural computations performed by RNN models against neural data.

Confirmed speakers:

Computation across scales in brain and beyond
Organizers: Shervin Safavi, Roxana Zeraati

Description:

The brain operates across multiple interacting scales from the level of ion channels in individual neurons to the population of interacting neurons and even beyond, in connection with the rest of the body. Hence, to fully understand the mechanisms underlying brain function and ultimately behavior, we need to study both neural dynamics and neural computations across multiple scales of organization. Last year, we organized a successful satellite workshop addressing the multi-scale dynamics of the brain (“Multi-modal understanding of brain and behavior”). This year, we propose a complementary workshop to discuss brain computations across different scales.

In this workshop, we bring together a broad group of scientists in computational and experimental neuroscience as well as artificial intelligence (AI), studying distinct types of computation across different scales. Our speakers will discuss how they apply various computational frameworks, including efficient coding, reinforcement learning, Bayesian inference, information theory, and deep learning, to formalize computational and learning algorithms not only from the level of single neurons to interacting brain areas, but also considering computations with non-neuron cell types (e.g., astrocytes), and eventually brain-body interactions. We aim to stimulate discussions on the benefits and challenges of different computational frameworks, how they can facilitate a multi-scale understanding of brain computation, and their potential applications to advancing AI algorithms. We believe that through this exchange of ideas, we can find new paths toward systemizing an approach to uncover a “hierarchy” of computation in the brain. 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 multi-scale brain research to gain a more holistic understanding of brain computation and cognition.

Confirmed speakers:

Emerging models in the evolution of neural systems and circuits
Organizers: Fred Wolf, Claudia Fichtel

Description:

Brains are among the most complex systems constructed by biological evolution. Understanding the principles of their design as versatile and efficient computational devices, reconstructing the paths through which the enormous diversity of animal nervous systems has emerged, and uncovering how the evolution of brains impacts animal behaviour are the ultimate goals of research into biological intelligence. Recent years have witnessed spectacular advances in our ability to uncover these fundamental evolutionary processes and principles. The question, however, of how selection and genetic drift, key innovations, and ecological niches shape the design of neural systems has remained notoriously hard to address. The Bernstein Conference satellite workshop “Emerging models in the evolution of neural systems and circuits” will provide a stage to critically assess key challenges and discuss novel approaches to dissect the evolutionary transformation of behaviour, cognition, and neural circuit function. Lightfoot, Fichtel, Huber, and Martelli will present study designs that bridge the gap between adaptive behaviour in natural habitats and neural circuit laboratory studies. They offer research approaches, spanning the whole range of nervous system complexity in bilateria, from nematodes to primates, that promise to directly link Darwinian fitness to the evolutionary modification of circuits and behaviour. Sachse, Vogel, and Engelken will present novel models, ranging from new model species to in-vitro and computational models, designed to address key questions in neural circuit evolution. As a whole, the symposium is set up to provide an inspiring and comprehensive landscape for discussing opportunities and challenges for decisive progress at the frontier of evolutionary biology and computational neuroscience.

Confirmed speakers:

Interaction of Synaptic Plasticity and Network Structure: Integrating Experimental and Multi-scale Modeling Approaches
Organizers: Jannik Luboeinski, Han Lu

Description:

The workshop overviews the critical features of empirically observed long-term neural plasticity and aims to trigger a discussion about how their interplay contributes to neural function and disease, founded on computational approaches at the microscopic and mesoscopic levels.

The workshop begins with an experimental perspective on synaptic dynamics, elucidating the intricate biological aspects of synaptic and structural plasticity across spatial and temporal scales. In the following, we dive into the modeling of microscopic dynamics, aiming to bridge the gap between theoretical frameworks and empirical observations, and to understand how structural and functional changes at the cellular level influence network-level phenomena. Then, we take a closer look at emergent network dynamics that can arise from functional and/or structural synaptic plasticity as well as given network topologies, seeking to examine the impact of the interplay between neuronal connections, activity patterns, and plasticity mechanisms on neural functions and disorders, which shall end in an open discussion.

Confirmed speakers:

Maximilian Lenz
Michael Faut
Stefan Rotter
Sandra Diaz Pier

Multisensory integration along the sensorimotor pathway
Organizers: Jan Clemens, David Deutsch, Marion Silies

Description:

Real-world experiences typically involve multiple senses. From social communication and taste perception to balance control and spatial orientation, integrating information from various senses is crucial for sensory perception, action selection, and motor control. In this workshop, we focus on a specific question regarding multisensory integration: What principles govern the organization of multisensory integration in the brain? Previous studies have demonstrated that multisensory integration can occur at different stages along the sensorimotor axis: early in sensory areas or late in associative and motor areas. However, it remains unclear what factors determine the point of convergence: the integrated sensory modalities, the behavioral context and task structure, or the characteristics of the resulting motor output? To discuss organizing principles of multisensory integration in the brain, we will bring together established researchers and young scientists who use a diverse range of model systems — flies, rodents, and humans — and methods — functional imaging, brain-wide calcium imaging, electrophysiology, and computational modeling. All researchers use systems-level and/or computational approaches to identify where and how multiple senses are integrated in the brain. To promote exchange among scientists from diverse backgrounds, talks are organized into two short sessions, each featuring speakers who employ different model systems and methods. The talks themselves will be kept concise and conceptual, with ample time reserved for discussion between talks and at the end of each session.

Confirmed speakers:

Neural Diversity and Computation – Towards a Mathematical Account of Tissue Heterogeneity in the Brain
Organizers: Richard Gast, Jeremie Lefebvre

Description:

Recent technological developments in cell type identification, synaptic tracing, and multi-neuron recordings have led to massive data sets for cell-type-resolved brain structure and dynamics. These data sets have not only identified new cell types, but also revealed a previously underappreciated level of neural and synaptic heterogeneity within cell types. In contrast, most mathematical models of brain function focus on networks with a few classes of neurons (e.g., one excitatory and one inhibitory), where neurons within a particular class are considered as identical. This workshop will address how different aspects of brain heterogeneity can be expected to affect neural dynamics and computation. Together, we will discuss (i) how the heterogeneity gap between neural network models and biological neural networks may bias our theories of structure-function relationships in the brain, and (ii) how various aspects of brain heterogeneity may interact to achieve stable brain function. In the first session, we will gain insight into the wealth of synchronization patterns that can arise in complex networks and how they are affected by structural heterogeneity. We will then learn about the level of neural heterogeneity expressed in the human cortex, and discuss how the results from complex network studies may translate to brain networks. In the second session, we will move on to discuss different approaches that have been employed to explain how neural heterogeneity supports computations in spiking neural networks, supported by evidence from computation with heterogeneous neuromorphic devices. In the final session, we will discuss a range of theoretical results that connect different heterogeneity-inducing aspects of neural networks to emergent network dynamics and computation. We will conclude the last session with an extended discussion of how the various heterogeneities that exist in the brain may serve distinct purposes in the organization of the brain dynamics and function.

Confirmed speakers:

Neuromorphic Tug-of-War: Balancing the Pull of Neuroscience and AI
Organizers: Matteo Saponati, Laura Kriener, Melika Payvand, Benjamin Grewe

Description:

Historically, Artificial Intelligence (AI) research has been inspired by understanding how the brain works and emulating its functioning principles with artificial machines (Mead 1990). Yet, current breakthroughs in AI are driven by Machine Learning (ML) algorithms that show unbelievable performances with only minimal inspiration from actual biological networks. This success has nudged the field of Neuromorphic Computing towards ML principles, with increasing hesitation in implementing additional biological mechanisms in artificial learning systems. Nonetheless, this trend has been counterbalanced by groundbreaking discoveries in Neuroscience, indicating novel learning mechanisms that offer fast, parsimonious, and potentially powerful solutions for credit assignment. This tug-of-war between Neuroscience and AI has centered on two crucial questions: Which biological features provide promising solutions for hardware implementation, and which are merely incidental, resulting from their biological substrates? Conversely, what could Neuromorphic Computing gain by integrating knowledge from the ML community, regardless of its biological plausibility? This workshop will bring together researchers in Computational, System, and Cellular Neuroscience, Neuromorphic engineering, and ML to discuss these conflicting perspectives: (a) What are the recent Neuroscience breakthroughs in understanding learning and intelligence? Which ones have not been explored in artificial systems yet? (b) What can we learn from the successes and failures of ML models?, and from a wider angle, (c) How much biological inspiration is needed for an artificial system to be “Neuromorphic”?

Confirmed speakers:

Physics of cognition: dynamical models of the ordered and disordered brain
Organizers: Hamidreza Jamalabadi

Description:

Neuroscience has made remarkable progress in recent years, significantly deepening our understanding of the brain’s relationship with behavior and its disorders. Despite these advancements, our capability to develop constructive models for cognitive functions like memory and emotion, which would allow for personalized, goal-directed interventions, is still in its infancy. This raises a crucial question: How close are we to establishing foundational models of cognitive functions that mirror the fundamental principles seen in physics and engineering, such as Newton’s laws of motion? Is this necessary or even plausible? More pressing is the question of how far we have come in leveraging our neuroscientific knowledge to develop concrete, computational solutions for clinical and other applications. A notable example is the limited application of functional neuroimaging in routine psychiatric care, despite substantial investments in this area. How far are we from utilizing neuroscience in psychiatry to shape treatment policies or create predictive models? Related, there is an emerging disconnect between the advancements in artificial intelligence (AI) and neuroscience evidence. For instance, the latest generations of artificial neural networks, like transformers, have shown improved performance but are increasingly deviating from principles grounded in neuroscience. Does this mean that progress in AI is becoming independent of neuroscience? Furthermore, it poses the question of whether two fundamentally different cognitive systems—one based on the human brain and the other on AI—can exist, each following its own unique construction principles, and if so, to what extent it is possible to port insights from biological brains into artificial brains, and vice versa? Faced with these challenges, the scientific community is increasingly recognizing the potential of complexity science, including generative AI and Dynamical Systems Theory (DST). These approaches are not new to neuroscience, but applying them to understand cognitive processes and disorders is a relatively recent development. It is only in the last few years that we have started to see generative models capable of realistically reconstructing memories—or at least creating qualitative models of them. Similarly, models tracking the progression of psychiatric symptoms over time and mathematical tools capable of deducing entire dynamical systems from partial observations subject to stochastic fluctuations are just emerging. This shift towards complexity science and advanced modeling techniques is driven by several factors. Our struggles to apply neuroscience knowledge practically, advancements in DST like data-driven reconstruction algorithms, the availability of big open datasets, and significant improvements in computational power have all contributed. In this workshop, our collective goal is to discuss the questions raised above around the physics of cognition and critically evaluate the potential and limitations of current methodologies from the perspective of dynamical system theory. We have gathered an exceptional ensemble of neuroscientists, coming from varied fields such as engineering, physics, and medicine, all united by a common interest in cognitive modeling. Our speakers stand out for their renowned contributions to neuroscience, employing and advancing computational techniques within the realms of machine learning, dynamical systems, control theory, and deep learning-driven generative AI. They apply these advanced methodologies to a broad spectrum of data types, including BOLD fMRI, electrophysiology, EEG, and behavioral data, to explore various cognitive functions and disorders. The discussions at our workshop will span a wide array of subjects, from understanding fundamental cognitive functions like memory to investigating changes in neurological conditions such as Alzheimer’s disease and psychiatric disorders like major depressive disorder. Our ultimate goal is to foster a rich dialogue that can lead to the identification and establishment of fundamental principles in the physics of cognition. These principles aim to bridge our understanding of both healthy and disordered cognition, enhancing the field of translational neuroscience. By doing so, we hope to make a meaningful contribution to neuroscience, bridging theoretical insights with practical applications. Our ambition is to advance the field significantly, improving how we address and manage neurological and psychiatric conditions. In pursuit of this vision, our aim is to stimulate discussions among the speakers and participants on:

Mathematical and Computational Tools for Studying Cognitive Processes: We are interested in exploring the types of mathematical and computational tools that are crucial for studying cognitive processes, especially when dealing with partial (temporal and spatial) observations subject to stochastic fluctuations. We will discuss the current state of the art and identify the most pressing next steps in this area.
Modeling Cognitive Functions: Our discussions will also focus on our current capabilities in modeling cognitive functions such as emotions, memory, and decision-making. We will examine the latest advancements for individual interventions and deliberate on the future direction of this research.
Dynamical Models in Neurological and Mental Disorders: Another key topic will be the current status of dynamical models in understanding and addressing neurological and mental disorders. This discussion aims to evaluate where we stand and what the future holds for this domain.
Building First Principles for Cognitive Processes: Lastly, we will consider the possibility of establishing first principles for the computational machinery underlying cognitive processes. This involves discussing the potential and limitations of developing a mathematical theory that can serve as a foundational framework for cognitive and computational neuroscience.

We are confident that this workshop, with its unique design, will captivate the interest of attendees from the Bernstein Conference, fostering lively discussions among computational and experimental neuroscientists from a variety of disciplines. It is our intention to facilitate a rich exchange of ideas across diverse fields, contributing to the shaping of future interdisciplinary collaborative efforts. This collaboration encompasses the realms of psychology, physics, engineering, computer science, and medicine. We strongly believe that the path to future breakthroughs and practical applications hinges on the synergy between experimental, clinical, computational, and engineering researchers. Our aim is to create a nurturing environment that promotes such interdisciplinary dialogue, which is essential for sparking significant progress in the field.

Confirmed speakers:

The neurobiological mechanisms of sequence processing
Organizers: Petros E. Vlachos, Alessio Quaresima, Hartmut Fitz

Description:

Recognition and generation of temporal sequences are fundamental capacities of human and animal cognition. Nevertheless, our understanding of the computational basis of these capacities remains limited. Many models have been proposed, but what do they tell us about the neural mechanisms? Does a unifying principle exist, or is sequence processing implemented by diverse neural algorithms? The workshop brings together theoretical and model-based approaches to explain sequence processing in the brain. It aims to foster discussion about the computational principles that support both the sequential propagation of neuronal activity (sequence generation) and the integration of temporal patterns (sequence recognition) in biological networks.

Confirmed speakers:

Understanding Local Third Factors in Synaptic Plasticity: Mechanisms and Functional Implications
Organizers: Andreas Schneider, Fabian Mikulasch, Viola Priesemann

Description:

Synaptic plasticity, the cornerstone of learning and memory, traditionally emphasizes the interplay of pre- and post-synaptic activity. However, recent research highlights the crucial role of “third factors” in shaping this process. These factors, distinct from pre- and post-synaptic activity, can be broadly categorized as “global” (e.g., dopaminergic reward signals) or “local” (e.g., postsynaptic membrane potential, or glial influence). While the importance of global third factors is well-established, the mechanisms and functional implications of local third factors remain largely unexplored. This workshop delves into this exciting frontier, exploring the intricate interplay between mechanisms and functions of local third factors in learning. We will bring together experts in biological learning, bio-inspired AI, and neuromorphic computing for a comprehensive discussion on key open questions. These questions encompass the biophysical underpinnings of local third factors, their impact on synaptic plasticity at the cellular and molecular level, and how they contribute to shaping network function and information processing across various brain regions. The workshop will conclude by discussing in which brain regions and in which learning tasks local third factors play a crucial role.

Confirmed speakers:

Bernstein Conference 2024

Photographic Retrospect

Invited Talks

Contributed Talks

Satellite Workshops

Folgen Sie uns auf