Bernstein Conference 2021

Location

Online

Date

Sep 21 – 23

Abstracts

Invited Lectures

The 2021 online Bernstein Conference focussed on the next generation in computational neuroscience. Therefore we give the floor to the new and rising names in the field.

Athena Akrami | University College London, UK
Formation and update of sensory “prior distributions” in working memory and perceptual decision making tasks

Timothy Behrens | University of Oxford, UK
Representing the structure of problems in the frontal hippocampal circuitry

Nadine Gogolla | Max Planck Institute for Neurobiology, Martinsried, Germany
Disentangling the neural basis of emotion

Guillaume Hennequin | University of Cambridge, UK
Failing to prepare is preparing to fail: a network theory of movement preparation and execution

Jennifer Li & Drew Robson | Max Planck Institute for Biological Cybernetics, Tübingen, Germany
A dynamical systems view of neuroethology: uncovering stateful computation during zebrafish foraging

Scott Linderman | Stanford University, USA
Finding sequences in neural spike trains with Neyman-Scott processes

Ashok Litwin-Kumar | Columbia University, New York City, USA
Generalizing theories of cerebellum-like learning

Ida Momennejad | Microsoft Research, New York City, USA
Multi-scale Predictive Representations & Human-like RL

Cristina Savin | New York University, New York City, USA
Task-specific routing of information in neural circuits via structured noise

Marion Silies | Johannes Gutenberg University, Mainz, Germany
Strategies for vision in dynamically changing environments

Daniela Vallentin | Max Planck Institute for Ornithology, Seewiesen, Germany
Neural control of vocal interactions in songbirds

Valentin Braitenberg Award Winner

Eve Marder | Brandeis University, Waltham, USA
Perturbations Reveal that Degenerate Circuits Hide Cryptic Individual Variability

Contributed Talks

Badr Albanna |University of Pittburgh, Pittsburgh PA, USA
Distinct synaptic plasticity mechanisms determine the diversity of cortical responses during behavior

Manuel Beiran | École Normale Supérieure – PSL University, Paris, France
Learning parametric models of cognitive tasks through low-dimensional neural manifolds

Yul HR Kang |University of Cambridge, Cambridge, United Kingdom
Spatial uncertainty provides a unifying account of navigation behavior and grid field deformations

Anna Kutschireiter | Harvard Medical School, Boston MA, USA
A Bayesian Perspective on the Fruit Fly’s Internal Compass

Claire Meissner-Bernard | Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland
Co-tuned, balanced excitation and inhibition in olfactory memory networks

Alireza Modirshanechi | Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
Novelty drives exploration, whereas surprise modulates learning: the interaction of novelty, surprise, and reward in human behavior

Torben Ott | Washington University in Saint Louis, USA
Near-optimal time investments under uncertainty in humans, rats, and mice

Ahnaf Ashhab Pathan | Quantitative Neuroscience Lab, Systems Cortex, Dhaka, Bangladesh
Fully topographic deep artificial neural network for reconstructing functional heterogeneity and internal representations of visual cortex

Our Community

Watch what the community says about the Bernstein Conference.

Talk Collection 2021

Here you’ll find invited and contributed talks from the Bernstein Conference 2021 online.

Satellite Workshops

Brain Criticality: The Past, the Present and the Future
Organizers: Dante Chialvo, Claudius Gros

Abstract:

The animals‘ adaptive behavior depends on their brain’s ability to sustain a large repertoire of spatio-temporal activity patterns. A fundamental problem in neuroscience is to understand the mechanisms by which brains can produce in a robust and flexible manner a huge range of neuronal dynamical configurations. A plausible solution, coming from statistical physics, invokes the emergence of complex phenomena exhibited universally by dynamical systems poised near a critical point of a second-order phase transition. In the last decade, work on this so called „brain criticality hypothesis“ attracted attention across disciplines.

The aim of the workshop is to provide a discussion forum for the perspective of the field: where do we stand, which are open issues, and where to go from here. The emphasis is not on technical presentation, but on a controversial assessment of the field. The workshop is connected to the „Focus on Criticality and Complexity in Brain Dynamics„, edited by the workshop organizers.

Speakers:

Control mechanisms for contextual computations and behavior
Organizers: Lorenzo Fontolan, Carlos Stein Brito

Abstract:

Throughout evolution, the nervous system has developed the ability to not only assign the appropriate responses to certain sensory stimuli, but also to rapidly switch to a different behavioral outcome according to context. Contextual information can be static, as in the encoding of a given set of relationships between items, or vary in time, as for the sense of urgency when we are late to the train station. To enable flexible, context-dependent decision-making, the brain must implement specific computational processes that reflect contextual information and dynamically modulate the activity in the brain areas where decisions are formed. Flexibility in neural processing of sensory information occurs in a matter of seconds, an interval too short for modifications of anatomical connections between sensory and executive areas. Hence, information flow must be flexibly routed via rapid changes in the neural communication between areas.

The goal of this workshop is to discuss novel mechanisms of contextual control, with focus on fast reconfiguration of functional connectivity via circuit dynamics, neuromodulators and inhibitory pathways.

Confirmed speakers:

Distributed Representation in RL: from AI to Dopamine
Organizers: Paul Masset, Pablo Tano

Abstract:

The temporal difference (TD) theory of dopaminergic activity in the basal ganglia has been one of the most influential ideas in neuroscience during the last two decades. It remains one of the few examples where a normative computation has been assigned to a genetically defined cell type. However, recent experimental work expanding the anatomical locations of recordings and the task designs has shown heterogeneity in dopamine responses that is not readily explained within the canonical TD framework. On the other hand, recent progress in AI has shown that extending the TD learning rule allows agents to learn more powerful representations that lead to improved performance in various domains. For example, improved performance has been obtained by extending the TD rule to learn the entire value distribution of states (instead of just their expectation), the expected values of states with multiple temporal discounts (instead of a single one) and the dynamics of transitions in the environment.

Confirmed speakers:

Frequency-preference responses to external perturbations: from neurons to networks
Organizers: Rodrigo F.O. Pena, Horacio G. Rotstein

Abstract:

Neuronal oscillations have been observed at many interacting levels of neuronal organization in the brain. Several structures process information in a frequency-dependent manner. Often frequency-dependent processing occurs not only at the circuit level, but also at the level of neurons and synapses, which possess frequency-filtering properties. Neuronal resonance generically refers to the ability of neuronal systems to exhibit an amplified response to oscillatory inputs at a preferred frequency band, and is often associated with preferred frequency phase responses. There is evidence that neuronal populations can maintain selective communication through oscillatory synchrony and that cognitive functions are flexibly chosen through such a mechanism.

The goal of this workshop is to gather theoreticians and experimentalists who work on these issues from different perspectives and focus on different levels of organization and the interactions between them. The topics will cover, but are not limited to, subthreshold, spiking and network resonance, oscillatory synchrony, communication through coherence, entrainment, cognitive flexibility, cross level interactions.

Confirmed speakers:

Heterogeneous distribution of synaptic weights: experiments and theory
Organizers: Francesco Fumarola, Lukasz Kusmierz

Abstract:

Numerous experimental findings have pointed out that correlates of synaptic efficacy are distributed in a heterogeneous manner, spanning a few orders of magnitude with many weak and very fewer strong synapses. However, how such heterogeneity emerges through development and plasticity, and what functional role it plays in neuronal circuits, is still not well understood. There is currently no common conceptual framework to study this aspect of neural organization, and most theoretical models rely on overly simplifying assumptions about the underlying synaptic weight distribution.
To fill this gap, we have invited experimental and theoretical neuroscientists to exchange recent results and ideas that link heterogeneity in synaptic weights, synaptic plasticity, neural dynamics and computation. The objective of this workshop is to bring together researchers who approach these links from diverse viewpoints and techniques. In doing so, we hope to encourage discussions and new collaborations, and to collate various models that concern the functional role of heterogeneity in synaptic weights. An important subgoal of the workshop is to devise robust experimental designs that can challenge the proposed theoretical models and interpretations.

Confirmed speakers:

Machine Learning meets Neuroscience: from Spikes to Stimulation
Organizers: Fabian Sinz, Alexander Ecker

Abstract:

Machine Learning, in particular deep learning, has become an important tool in computational neuroscience to bridge the gap between models and data. Recent experimental techniques in neuroscience yield an increasing amount of rich data that can be fitted with complex models using novel machine learning techniques. This provides a great opportunity to strengthen the link between complex theoretical models and data. For instance, not only have deep networks set new standards in predicting responses of neural populations to arbitrary stimuli and the synthesis of novel stimuli for experimental manipulation, but novel probabilistic machine learning techniques also help to fit parameters of simulation-based (spiking) models to data. At the same time, deep learning serves as an inspiration for how biological networks could learn complex problems using biologically plausible learning rules. With this workshop, we want to highlight the opportunities of modern machine learning in (computational) neuroscience by showcasing a diverse set of successful applications and discussing opportunities for how novel techniques could help test theoretical models with experimental data.

Confirmed speakers:

Maintaining function in the presence of ongoing change
Organizers: Sven Goedeke, Felipe Yaroslav Kalle Kossio

Abstract:

Recent experiments have revealed ongoing changes in the brain. In particular, synapses show intrinsic dynamics, even in the absence of activity. They also appear and vanish over time leading to a dynamic connectome. On a higher level, representations of information by neural activity exhibit continuous change or drift. In contrast, memories and behavior often remain stable for a long time. This workshop aims to bridge the different levels of ongoing changes and to discuss how function can be maintained in their presence. Specifically, the workshop will explore the following questions. How can unstable and intrinsically dynamic synapses support stable memories? Are ongoing changes of synapses a cause of representational drift? What are the characteristics of representational drift in different brain systems? What roles do homeostasis and redundancy play? How are the ongoing changes related to disorders and aging of the brain?

The workshop will bring together experimental, computational and theoretical perspectives on these questions. In several talks, researchers will present recent experimental data demonstrating drift or stability of neural representations. Furthermore, modeling studies addressing different aspects of ongoing changes will showcase current computational and theoretical approaches. There will be time to interactively discuss questions and ideas among the participants. We hope the workshop will foster the exchange between the different perspectives and aid our conceptual understanding of a dynamic structure-function relationship in neuroscience.

Confirmed speakers:

Recent Developments for Biologically Realistic Neural Mass Models
Organizers: Alessandro Torcini, Simona Olmi

Abstract:

The workshop will focus on recent developments concerning neural mass models, i.e., mean field models able to reproduce biologically realistic dynamics of networks of spiking neurons [1]. In particular, emphasis will be devoted on one side to mean field models for the balanced state, representing a fundamental aspect of cortical dynamics [2-6], and on the other side to next generation neural mass models, recently developed, and able to reproduce exactly the dynamics of spiking neural networks [7].

The workshop aims to favour a comprehension of recently developed neural mass approaches and an open discussion on their relevance and limits of applicability. In particular, a close comparison with experimental results will be performed. The topics will range from the emergence of long-lasting fluctuations and correlations in balanced cortical circuits [8-9], to theta-gamma nested oscillations [10-11], from short-term memory [12-13] to neural wave propagation [14].
The inclusion in the neural mass models of realistic aspects of neural dynamics, such as the presence of background noise, sparseness in the connection [15-16], adaptation [17] and synaptic plasticity [12-13] will also be addressed in detail during the workshop.

[1] Carlu et al. Journal of Neurophysiology, 123(3), 1042 (2020).
[2] C. van Vreeswijk and H. Sompolinsky, Science 274, 1724 (1996).
[3] A. Renart, J. de la Rocha, P. Bartho, L. Hollender, N. Parga, A. Reyes, and K. D. Harris, Science 327, 587 (2010).
[4] A. Litwin-Kumar and B. Doiron, Nat. Neurosci. 15, 1498 (2012).
[5] R. Rosenbaum and B. Doiron, Phys. Rev. X 4, 021039(2014).
[6] J. Kadmon and H. Sompolinsky, Phys. Rev. X 5, 041030 (2015)
[7] E. Montbro’, D. Pazo’, and A. Roxin, Phys. Rev. X5, 021028 (2015)
[8] F. Mastrogiuseppe and S. Ostojic, PLOS Computational Biology (2017)
[9] Dahmen, D., Gruen, S., Diesmann, M., & Helias, M. Proceedings of the National Academy of Sciences, 116(26), 13051-13060. (2019)
[10] M. Segneri, H.Bi, S. Olmi, A.Torcini, Frontiers in Computational Neuroscience , 14:47 (2020).
[11] A.Ceni, S. Olmi, A. Torcini, D. Angulo Garcia, Chaos , 30, 053121 (2020).
[12] Schmutz, V., Gerstner, W., & Schwalger, T. The Journal of Mathematical Neuroscience, 10(1), 1-32. (2020)
[13] H. Taher, A. Torcini, S. Olmi, PLOS Computational Biology, 16(12): e1008533 (2020)
[14] Ã Byrne, R.D. O’Dea, M Forrester, J Ross, S Coombes, Journal of Neurophysiology 123 (2), 726-742 (2020).
[15] M. di Volo, A. Torcini, Phys. Rev. Lett. 121, 128301 (2018)
[16] H. Bi, M. Segneri, M. di Volo, A.Torcini, 2, 013042 (2020)
[17] Di Volo, M., Romagnoni, A., Capone, C., & Destexhe, A. Neural computation, 31(4), 653-680. (2019)

Confirmed speakers:

The geometry of neural activity: low-dimensional dynamics and high-dimensional representations
Organizers: Joao Barbosa, Adrian Valente, Yuxiu Shao, Ljubica Cimesa

Abstract:

The dimensionality of neural representations is at the center of an intense debate. Theoretically, a low-dimensional representation buys you robustness and generalizability. On the other hand, high-dimensional representations are more flexible, because they are more easily decoded by downstream areas. This theoretical tension is reflected in paradoxical findings in neuroscience. With reconciliatory potential, some have argued that the widespread low-dimensional representations in neurophysiological recordings could simply reflect the dimensionality of the task the animals are engaged with.

This workshop will bring together evidence from both views, with the hope of clarifying their interaction in particular in three broad fields:
1. Decision Making
2. Learning in Artificial Neural Networks
3. Motor Learning

With the help of the audience and the speakers, we would like to address the following general questions:
1. Does the brain use both high and low dimensional representations, and if so, for what purposes?
2. Can the paradoxical empirical findings be reconciled? For example, seeing neural representations as low-dimensional with respect to the neural space (~10^10) but high dimensional with respect to the task variables (<10)?
3. How does dimensionality change during learning, both in biological and artificial networks?

Confirmed speakers:

The inner life of dendrites - mechanisms behind synaptic input integration
Organizers: Fabio Sartori, Surbhit Wagle, Tatjana Tchumatchenko

Abstract:

Synapses are the fundamental units of information transmission in the brain. Maintenance and change in the synaptic efficacy require precise regulation of proteins and mRNAs in the synapses, dendrites, and axons. This is a challenge because the typical half-life of a protein is only a few days, while the time-scale of the behavior the proteins regulate ranges from few minutes to several decades.
This workshop aims to explain this gap in time scales by discussing the mechanisms that allow the neurons to regulate the availability of relevant proteins and mRNAs across the neuronal processes. The speakers in the workshop will cover topics ranging from mRNA and protein turnover to dendritic transport and relate them to neuronal computation.

Confirmed speakers:

Bernstein Conference 2021

Invited Lectures

Valentin Braitenberg Award Winner

Contributed Talks

Our Community

Talk Collection 2021

Satellite Workshops

Folgen Sie uns auf