Brain Dynamics and Statistics: Simulation versus Data (17w5036)
Susanne Ditlevsen (University of Copenhagen)
Andre Longtin (University of Ottawa)
Priscilla Greenwood (University of British Colombia)
The data we have in mind come from intra- and extracellular recordings of single neurons, revealing both spiking and subthreshold patterns, local field potential data, spatio-temporal measurements which show us spatial patterns with the extra dimension of time, EEG and fMRI data. Specific questions are: What are the important modes of information transmission in the brain? Does it occur via spiking patterns, or time averages of patterns? Are various rhythms also important? Surprisingly little is known. We know about some simple aspects of the operation of primary visual cortex, such as how single cells code for edges or movement at different angles; but the representation of the many features that make up a stimulus across different spatio-temporal scales of activity is an unsolved puzzle.
In order to make progress we need to bring together people who gather neural data and understand lab work, people who have an extensive but usually specialized current knowledge about the brain, people who work in the mathematics of dynamical systems, particularly stochastic dynamical systems, and people who have facility in the statistics of stochastic processes. The common requirement is that these people are devoted to understanding the dynamics of the brain.
Despite a considerable recent body of literature on statistical inference and stochastic modeling in neuroscience, there remain substantial unsolved problems and challenges, some of which we will address during the workshop. Examples are: understanding the role of stochasticity in the brain, developing efficient Monte Carlo methods for inference, understanding the relationship between system behavior (i.e. bifurcations and their stochastic counterparts) and statistical properties of estimates based on data from these systems, and reconstructing the structure of network models from partially observed data.
The workshop has the potential to have large impact, if successful, by reshaping the way data is processed and understood in the biological sciences, from a dynamical stochastic approach. Neuroscience is experiencing an imbalance between prowess in data collection and deficiency in data interpretation. The improved models for data processing are expected to be of significant theoretical interest from biophysical, mathematical and statistical viewpoints. Our work has the potential to contribute to all parts of neuroscience, from single cell analysis to the understanding of neuronal network dynamics, and potentially beyond.
We will use statistical and methodological issues as a means to bring out common challenges and facilitate further cross-disciplinary collaboration. The meeting will begin with a short introductory session allowing each participant a few minutes to describe their own research interests as a means of facilitating informal discussion and collaboration, and the final day will be reserved for discussions of important future problems in the field.
We have contacted a number of prominent researchers within computational neuroscience, statistics and stochastic modeling with applications in neuroscience, all of whom expressed enthusiasm for participating in this workshop. The potential candidates who have confirmed their interest come from Canada, United States, Japan, Australia, France, Spain, Germany, United Kingdom, Italy, Switzerland, Czech Republic, and Denmark. A number of these people have an extremely high profile in this field. We will also invite a number of younger people at the level of postdoc and PhD.