Research

(see also: International School for Scientific Computing, ISSC )

The Stoop Group is a research group at the Institute of Neuroinformatics in Zürich (INI, part of ETH Zürich and University Zürich). The INI deals with the questions, of how biological systems compute, and how the knowledge gained about biological computation can be used to build technological devices.

The main research interests of the group can be classified under theoretical neuroinformatics: The consideration of questions like information representation, and computing in biological systems, and the application as possible answers for solving technological problems (e.g. noise cleaning); lie at the center of the groups interests. The theory of dynamical systems, information theory and statistical physics form the theoretical basis of the group's research.

The Group Leader is Prof. Dr. habil. Ruedi Stoop; mathematician and theoretical physicist. The group itself is a highly interdisciplinary group, that unifies competences in mathematics, theoretical physics, computer science, engineering, biology and philosophy.

Auditory modelling

Investigation of the information transfer in the auditory system. Design of computer simulation models of the basilar membrane motion in the cochlea of mammals, by a nonlinear dynamics approach. Electronic circuit implementation, miniaturization and possibly development of new hearing aids.

Previous and current research

• Development of a nonlinear (Hopf) cochlea model.
• Electronic implementation of the Hopf cochlea model.

Future research

• Simulation of different types of cochlea damage, comparisons of their performance with existing analog implementations (silicon cochlea).
• Application of tools for measuring the information loss generated by cochlea damage and the information recovery when using a hearing aid.

Dynamical systems theory and neuroscience

Determination of the generic mesoscopic properties of neocortical networks, and of communication and coding strategies in biological neural networks. Characterization of efficiency of cortical information processing by means of entropy. Determination of the role of noise in biological neural networks.

Previous and current research

As a consequence of previous research concentrating on simple pyramidal neurons, we are led to the following statements:

• For inhibitory binary interaction among neurons, chaotic response emerges on an open set of nonzero Lebesgue measure of the parameter space, for high coupling. This proves that in the brain, chaotic behavior is already introduced on a local level.
• Arnold tongues provide an efficient neural coding scheme. This coding unifies rate coding with phase coding.
• Synchronization of neurons in the brain requires generically strong dynamical sources of driving. For quasi-static networks, synchronization is only possible for very strong local inhibitory circuits.
• Interspike distributions of spiking neurons cannot be expected to be typically of Poissonian type. They are more accurately described by long-tailed distributions.

Future Research

• Extension of our mesoscopic investigations to other types of neurons.
• Investigation of the role of regular and of chaotic bursting for synchronization.
• Investigation (in simulations and in vitro) of the selectivity of different neuron types towards periodic discharge patterns.
• Extension of insight from idealized towards more realistic networks that take into account the properties of an increasing number of neuron types.
• In vivo-testing of locking as a possible coding scheme.
• Analyzing the effects of connectivity on synchronization.

Data and pattern analysis

Development of tools and algorithms for data analysis, noise-reduction and for biologically motivated pattern recognition based on dynamical systems theory and statistical physics.

Previous and Current Research

• Noise-cleaning methods for time series data of chaotic origin.
• Statistical analysis of neuron inter-spike-interval data (alpha-stable distributions).
• Development of spin-model-based tools for clustering.
• Development of a measure of the complexity of time series based on the thermodynamic formalism.
• Development of correlation-integral based methods for pattern recognition in neuronal spike trains.

Future research

• Application of spin-model based clustering for classification of substances gained by combinatorial chemistry according to their therapeutical potential.
• Analysis of the temporal structure (spike pattern detection) of neuronal spike trains. Relate the appearance of spike patterns to computational processes in neurons.
• Application of the thermodynamic formalism based complexity measure to neuronal data.