CNS*2026 Workshop on Methods of Information Theory in Computational Neuroscience

Information in the brain. Modified from an original credited to dow_at_uoregon.edu (distributed without restrictions)

14th and 15th of July, 2026

Halifax, Canada

CNS*2026

Aims and topics

Methods originally developed in Information Theory have found wide applicability in computational neuroscience. Beyond these original methods there is a need to develop novel tools and approaches that are driven by problems arising in neuroscience. A number of researchers in computational/systems neuroscience and in information/communication theory are investigating problems of information representation and processing. While the goals are often the same, these researchers bring different perspectives and points of view to a common set of neuroscience problems. Often they participate in different fora and their interaction is limited. The goal of the workshop is to bring some of these researchers together to discuss challenges posed by neuroscience and to exchange ideas and present their latest work. The workshop is targeted towards computational and systems neuroscientists with interest in methods of information theory as well as information/communication theorists with interest in neuroscience.

Registration and Access

The workshop will be held as a part of the CNS*2026 in Halifax, Canada. Please see the CNS*2026 website for registration to the workshops (this is required to attend).

Organising committee

Speakers

The following are confirmed (and tentatively confirmed) invited speakers for the workshop. We will add contributed short talks closer to the event (as per below).

Call for contributed talks

Call for contributions is now open.

In addition to our invited speakers, we are now calling for contributed short talks. If you are interested in contributing a talk, please send a title and abstract to Joseph Lizier (joseph.lizier@sydney.edu.au) and Marilyn Gatica (marilyn.gatica@nulondon.ac.uk).

We will review submissions weekly starting May 25, and the call will remain open until all available slots have been filled.

Program

Tuesday 14th Wednesday 15th
Chair: Joseph Lizier Chair: Demian Battaglia
09:00 - 09:45 Marilyn Gatica
"Towards an integrated computational approach for diversity-sensitive personalized medicine"
Thomas Varley
"Why should we care about synergistic information?"
09:45 - 10:30 Raphaël Lafond-Mercier
"Parallel gradient computations of mutual information between stimulus sequences and adaptive responses"
Jaroslav Hlinka
"Information Dynamics in Complex Brain Networks: From Causal Inference to Higher-Order Structure"
10:30 - 11:00 Coffee Break
Chair: Thomas Varley Chair: Jaroslav Hlinka
11:00 - 11:45 Bratislav Misic
"Structure and computation in brain networks"
Simachew Mengiste
"Dynamic Functional Connectivity Resolves Brain Integration-Segregation Trade-off Under Costly Links"
11:45 - 12:30 Patricio Orio
"Emergence of High-order interactions in dynamical systems"
Leyla Roksan Caglar
"Same Compression Principle, Different Geometry: Rate-Distortion Signatures Dissociate Biological and Artificial Visual Systems"
12:30 - 14:00 Lunch Break
Chair: Marilyn Gatica
14:00 - 14:45 Joseph Lizier
"Constraints on information decomposition from a target chain rule"
14:45 - 15:30 Demian Battaglia
Heatwaves in the Brain"
15:30 - 16:00 Coffee Break
16:00 - 16:45 Discussion
TBA
16:45 - 17:30 Discussion
TBA

Abstracts

More to come soon!

Demian Battaglia - "Heatwaves in the Brain" Pleasant weather is entropic: sunny days alternate with rain, temperatures fluctuate, and gentle winds constantly reshape atmospheric conditions. By contrast, heatwaves are characterized by persistence. Once established, they tend to linger, day after day, reflecting a temporary trapping of atmospheric dynamics into unusually stable configurations. From the perspective of dynamical systems, these episodes correspond to a slowing of the system's exploration of phase space, reduced dynamical instability, and diminished production of information.
Recent advances in climate physics have exploited the statistics of extreme events to derive practical estimators of local dynamical stability, avoiding the notoriously data-hungry estimation of Lyapunov exponents. These methods quantify the persistence of trajectories in phase space through local recurrence statistics, providing experimentally tractable proxies for the system's instantaneous dynamical flexibility.
Here, we transfer one of these markers from climate science to neuroscience. Applying these stability estimators to high-density EEG recordings, we find that early Alzheimer's disease model mice exhibit an increased incidence of "neural heatwaves" -- or dynamic clogs -- i.e. transient episodes of abnormally persistent brain dynamics. These episodes emerge before classical electrophysiological biomarkers, are associated with deficits in associative memory, and are reversed by non-invasive 40 Hz sensory stimulation, which restores a more fluid exploration of neural state space.
Beyond Alzheimer's disease, this work suggests that cognition may critically depend not only on which brain states are visited, but on how fluidly the brain moves among them. Information processing may ultimately rely on maintaining the delicate balance between stability and instability that allows the continuous generation of novel neural configurations.

Leyla Roksan Caglar - "Same Compression Principle, Different Geometry: Rate-Distortion Signatures Dissociate Biological and Artificial Visual Systems"
Efficient coding theory predicts that biological perceptual systems compress sensory input optimally under resource constraints, with the systematic structure of errors reflecting the geometry of that compression. Here we operationalize this principle using rate–distortion theory (RDT) to characterize how any system - biological or artificial - trades representational fidelity for informational efficiency. Treating stimulus-response behavior as an effective communication channel, we infer rate–distortion (RD) frontiers directly from confusion matrices and summarize each system with three geometric signatures: slope (β), curvature (κ), and area under the RD curve (AUC), capturing the marginal cost, abruptness, and overall efficiency of the accuracy-compression trade-off respectively. Applying this framework to human psychophysical data and 18 deep vision models across 12 families of controlled image perturbations at graded severities, we find that both biological and artificial systems follow a common lossy-compression principle but occupy systematically different regions of RD space. Humans exhibit smooth, flexible trade-offs characteristic of near-optimal efficient coding, while deep networks operate in steeper, more brittle regimes even at matched accuracy, with geometry dissociable from performance across training regimes. Critically, behavioral RD signatures track internal representational geometry, evidenced by the behaviorally inferred compression structure correlating with internal representational dissimilarity across all models. Moreover, κ and AUC constitute complementary signatures: AUC tracks representational geometry independently of accuracy and captures the efficiency of learned categorical representations, with alignment strengthening specifically at deep categorical layers, while κ captures the abruptness of compression transitions at the level of the behavioral RD frontier and shows a distortion-type-dependent relationship with accuracy degradation dynamics. These results establish RD geometry as a compact diagnostic of percep- tual compression strategy that recovers mechanistically interpretable structure in internal representations from behavioral input alone and extends naturally to the direct characterization of compression geometry in neural population activity.

Marilyn Gatica - "Towards an integrated computational approach for diversity-sensitive personalized medicine"
TBA.

Nicolás Hinrichs - "Information Geometry for Inter-Brain Network Analysis in Hyperscanning"
Inter-brain synchrony during social interaction is increasingly studied via hyperscanning, yet standard synchrony measures collapse the rich geometric structure of neural co-variation into scalar indices. I will present HyPhi, a Python toolbox implementing information geometry and Forman-Ricci curvature for dual-EEG hyperscanning data. Treating inter-brain connectivity as a weighted network evolving over interaction time, HyPhi computes curvature profiles that distinguish categorically structured synchrony from noise-driven co-fluctuation, recovering mechanistically interpretable signatures inaccessible to mutual information or coherence alone. I will show results from live hyperscanning recordings, where inter-brain curvature indexes the degree to which two agents share a generative model of their interaction. The approach generalises to hypergraph representations of multi-brain dynamics and connects to broader questions about the geometry of collective inference.

Raphaël Lafond-Mercier - "Parallel gradient computations of mutual information between stimulus sequences and adaptive responses"
Mutual information (MI) is notoriously difficult to compute and its gradients with respect to model parameters even more so. Inspired by the encoding of time sequences in the brain of the weakly electric fish, we propose a few simplifications of the model that allows tractable computation of both quantities. While the computation remains expensive, we present two methods exploiting the parallel nature of the estimation of MI to retrieve its gradients in reasonable time. This is enabled using modern hardware such as GPUs to process large amounts of samples concurrently. We then apply this machinery in the context of encoding the time between encounter events in adaptive responses observed in the fish to retrieve optimal distributions of biophysical parameters through gradient ascent. We show from first principles that a response threshold enables populations of neurons to tile the stimulus prior, relating recovery time constants with time intervals found in nature. Because thresholds keep neurons silent outside of their responding range, the uncovered code is efficient in both number of neurons and in firing rate.

Joseph Lizier - "Constraints on information decomposition from a target chain rule"
TBA.

Simachew Mengiste - "Dynamic Functional Connectivity Resolves Brain Integration-Segregation Trade-off Under Costly Links"
Dynamic functional connectivity (dFC) is ubiquitously observed in the brain, but why functional networks should remain dynamic even at rest is unclear We asked whether temporal reconfiguration becomes advantageous when keeping a functional link active is costly. Modeling resting-state dFC as a temporal communication network, we show that empirical dFC outperforms equal-cost static architectures by increasing the reach and speed of information spreading in sparse regimes. Unlike more randomized temporal null models, however, it also preserves strong local cohesiveness, temporal clustering, rapid return of information to its source, and high neighborhood retention. Empirical dFC therefore achieves a compromise between large-scale integration and transient local segregation. This compromise is not explained by generic temporal variability, nor by partially frozen null models with persistent templates. A connectome-based mean-field model reproduces several key features, including high spatial and temporal clustering and strong integrative and segregative performance, but remains more stable over time than the empirical data. Our results indicate that empirical dFC reflects a structured regime of controlled persistence and renewal, in which local neighborhoods are maintained long enough to support transient recirculation before broader network-wide spreading occurs. Dynamic functional connectivity thus appears to be a resource-efficient solution to competing communication demands.

Bratislav Misic - "Structure and computation in brain networks"
The brain is a complex network of anatomically connected and functionally interacting neuronal populations. The wiring of the network allows its components to collectively transform signals representing internal states and external stimuli. Recent technological and analytic advances provide the opportunity to comprehensively map, image and trace connection patterns of nervous systems in multiple species, yielding high-resolution connectomes of individual brains. Yet how computation and functional specialization emerge from network architecture remains unknown. In this talk I will discuss how an area of artificial intelligence - reservoir computing - can be used to design bio-instantiated artificial neural networks, and to study the link between structure, dynamics and computational function. 

Thomas Varley - "Why should we care about synergistic information?"
The phenomenon of higher-order "synergistic" information has become a source of widespread interest in neuroscience, information theory, and the study of complex systems. Various proposals have identified synergy with: emergent properties, information integration, and even phenomenological consciousness. A plethora of statistical approaches have been introduced, all designed to wring synergistic information out of multivariate datasets. Despite this attention, the specific relevance of synergy to neural and cognitive processes remains abstract. Unlike redundancy, which readily maps onto notions of synchrony and coherence, it is less obviously clear what the presence (or absence) of synergistic information in a dataset in telling us. Building on prior links between synergistic information and information modification, this talk argues that synergistic information is a statistical fingerprint of task-relevant "computation" in the brain, occurring when signals from multiple functional systems interact to solve some kind of task. We review evidence from neuroscience and artificial intelligence research that connects synergy with successful task performance, as well as the mathematical links between synergy and causal colliders in the theory of causal inference. This perspective suggests that the identification of synergistic dependencies in the brain (and other systems) may be of practical, as well as theoretical, relevance to cognitive and clinical neuroscience.

Previous workshops

This workshop has been run at CNS for over two decades now -- links to the websites for the previous workshops in this series are below:

  1. CNS*2025 Workshop, July 8-9, 2025, Florence, Italy
  2. CNS*2024 Workshop, July 24, 2024, Natal, Brazil
  3. CNS*2023 Workshop, July 18-29, 2023, Leipzig , Germany
  4. CNS*2022 Workshop, July 19-20, 2022, Melbourne, Australia
  5. CNS*2021 Workshop, July 06-07, 2021, Online!
  6. CNS*2020 Workshop, July 21-22, 2020, Online!
  7. CNS*2019 Workshop, July 16-17, 2019, Barcelona, Spain.
  8. CNS*2018 Workshop, July 17-18, 2018, Seattle, USA.
  9. CNS*2017 Workshop, July 19-20, 2017, Antwerp, Belgium.
  10. CNS*2016 Workshop, July 6-7, 2016, Jeju, South Korea.
  11. CNS*2015 Workshop, July 22-23, 2015, Prague, Czech Republic.
  12. CNS*2014 Workshop, July 30-31, 2014, Québec City, Canada.
  13. CNS*2013 Workshop, July 17-18, 2013, Paris, France.
  14. CNS*2012 Workshop, July 25-26, 2012, Atlanta/Decatur, GA, USA.
  15. CNS*2011 Workshop, July 27-28, 2011, Stockholm, Sweden.
  16. CNS*2010 Workshop, July 29-30, 2010, San Antonio, TX, USA.
  17. CNS*2009 Workshop, July 22-23, 2009, Berlin, Germany.
  18. CNS*2008 Workshop, July 23-24, 2008, Portland, OR, USA.
  19. CNS*2007 Workshop, July 11-12, 2007, Toronto, Canada.
  20. CNS*2006 Workshop, June 19-20, 2006, Edinburgh, U.K.

Image modified from an original credited to dow_at_uoregon.edu, obtained here (distributed without restrictions); modified image available here under CC-BY-3.0