Complex Nonlinear Neural Dynamics

CNS 2001 Workshop


The Workshop on Complex Nonlinear Neural Dynamics at the CNS 2001 conference was organized by P Andras (U Newcastle, UK), R Kozma (U Memphis, TN),  A Assadi (U Wisconsin, WS), D DeMaris (U Texas, TX), and T Wennekers (MPI-MIS, Leipzig, Germany). The keynote speaker of the workshop was Professor WJ Freeman (Berkeley, CA).

Brief evaluation and future plans:

The workshop took place on July 5, 2:30 pm - 5:00 p.m. The workshop is considered a success with approximately 35 attendees and active interaction between speakers, panelists, and audience. It has been decided to continue the interaction between the interested attendees in the form of a web site and associated discussion group, as well as future joint activities like conference workshops, special sessions, etc.

Summary of the discussions:

Professor Freeman gave an insightful talk about his theory on the role of neural chaos. He described the role and functioning of mesoscopic neural processes. These serve as the main underlying mechanisms for the emergence of attractors and switching between attractors in the space of spatio-temporal neural activity. His talk provided an overall view on where and how such attractor dynamics might be detected in the brain, with special emphasis on the olfactory and visual nervous systems. He also presented a brief review on the emergence and disappearance of attractor states, emphasizing the role of stochastic noise in these processes.

The keynote speech was followed by two panel discussions.


The first panel consisting of Professor WJ Freeman, Dr R Kozma, Dr D DeMaris and Dr T Ferree (U Oregon, OR) discussed the measurement and interpretation of complex dynamics in biological neural networks. Each participant presented briefly his view on topic - related issues. Professor Freeman questioned the validity of time ensemble averaging measurements, pointing to the fact that these averaged measurements in some cases may reveal only the underlying noise instead of the meaningful spatio-temporal signal. He proposed the use of spatial ensemble averaging instead to analyse the spatial patterns of broadband 'chaotic' carrier waves and discard the spatially incoherent impulse response to perturbation. Dr Kozma presented his theory about the role of noise in dynamical neural systems, and discussed measurement issues about stochastic noise and complex nonlinear signals. He described the use of low level noise corresponding to basal unit activity that maintains chaotic attractors and enables the state transitions. He also described a new computational method based on chaotic dynamical memories. Dr Ferree presented briefly his work on measurement of chaotic nature in time series, emphasizing the multi-fractal nature of some natural time series, in particular in EEG measurements. Dr DeMaris discussed the use of simple dynamical systems as models of brain dynamics, and addressed issues about coupled logistic maps and the use of symbolic dynamics in the analysis of them.

The second panel consisted of Professor WB Levy (U Virginia, VA), Dr T Wennekers, Dr P Andras and Dr A Assadi. They discussed the use of complex neural dynamics in the context of neural computation. Professor Levy presented briefly the work of his team on a simple dynamical model of the memory, and discussed the need for controlling the emerging chaos in such systems. Dr Wennekers discussed tentative functional roles for complex dynamics focussing on signatures of chaos like irregularity, broad-band power spectra, strange and multiple attractors, and their sensitivity against parameter changes. Dr Andras presented the main ideas of a new approach to model and analyse neural dynamics in a computational context. He suggested that attractors forming in the spatio-temporal activity space of small neural assemblies may be used as basic computational units. Dr. Assadi discussed a mathematical approach to the interpretation of complex nonlinear dynamics in computational context. His approach is rooted in statistical learning theory and leads to an asymptotic theory of computational neural dynamics.

In both cases the brief presentations of the panellists were followed by a discussion of the addressed topics.




Please send comments to .


This web page was created on August 28, 2001. Last update: August 28, 2001.