Page Content

Research Group "Modelling of dynamic perception and action"

Research Group Leader

What we do

Our group develops (i) computational models for cognitive processes like decision making, recognition and learning of complex spatiotemporal patterns, dynamic multisensory integration, spatial navigation and motor control, and (ii) neuronal models at several spatiotemporal scales, e.g. neural mass or single neuron models. For specific experimental neuroscience findings, we derive mathematical and functional mechanisms and aim at providing a mathematical framework for cognitive or neuronal processes.

Methods

As a modelling framework, we mostly use Bayesian online inference for nonlinear dynamical systems. Using variational inference schemes, we construct neurobiologically plausible systems that receive dynamical sensory input and can learn and recognize this input and act accordingly, in an online fashion. This enables us to model brain function as predictive and self-organising dynamics operating at multiple time-scales.

Areas of interest

We model phenomena in several neuroscience topics like auditory and audiovisual communication, multisensory integration and plasticity, birdsong generation and recognition, decision making, spatial navigation, novel applications of recurrent neuronal networks, and dendritic computation.

Experiments

We collaborate with several experimental groups to develop cognitive and neuronal models. Experiments are typically done using functional magnetic resonance imaging, electro/magnetoencephalography, psychophysics, local field potentials, and two-photon laser microscopy.

Positions

Currently a Master thesis project is available.

Our Group

From left to right: Stefan Kiebel , Dimitrije Markovic, Hame Park, Sebastian Bitzer , Burak Yildiz , Jelle Bruineberg, not on picture: Kai Dannies.

Group members



Current projects


Generation and Recognition of Bird Songs
How do birds sing complex and elegant melodies? This question has kept researchers busy for decades and still we don't have a complete understanding of the mechanisms underlying birdsongs. In this project, we use dynamical systems theory to propose a biologically plausible, hierarchical model for birdsong generation and introduce a novel Bayesian inversion scheme for online recognition of birdsongs. read more
Recognizing Recurrent Neural Networks
Recurrent Neural Networks (RNNs) model dynamic processing in the brain. We reinterpret standard RNNs as generative models for dynamic stimuli and show that 'recognizing RNNs' resulting from Bayesian inversion of these models combine many important aspects of brain function. read more
Audiovisual speech perception
When we listen to a speaker we understand the conveyed message better when we also see the speaker's face. There is strong experimental evidence that this audiovisual integration of speech relies on specific brain mechanisms. However, there are currently no compelling computational models to explain key experimental findings. In this project we aim at developing a neurobiologically plausible model of audiovisual speech perception and testing its predictions using imaging techniques, in collaboration with the group of Katharina von Kriegstein. read more
Decision making and active Inference
Using a novel mathematical framework for action and decision making [4], we aim at developing artificial agents that close the loop between brain and environment. These agents, and their emerging actions and decisions, can be used to test theoretical predictions using neuroimaging experiments. In collaboration with John O'Doherty and Jean Daunizeau. read more
Spatial Navigation
Using nonlinear dynamical systems, we model spatial navigation and path learning as Bayes-optimal inference in a spatiotemporal environment, at multiple time-scales. We aim at modelling several experimental findings about the hippocampus and its interaction with cortical areas.read more
Free-energy and dendritic self-organisation
We pursue recent observations that, through selective dendritic filtering, single neurons respond to specific sequences of presynaptic inputs. We try to provide a principled and mechanistic account of this selectivity by applying the free energy principle to a dendrite that is immersed in its neuropil or environment. We assume that neurons self-organize to minimise a free energy bound on the self-information or surprise of presynaptic inputs that are sampled. In collaboration with Karl Friston. read more

Collaborations

Karl Friston Wellcome Trust Centre for Neuroimaging, London, UK.
Jean Daunizeau Brain and Spine Institute (ICM), Paris, France.
Katharina von Kriegstein, Max Planck Institute for Cognitive and Neuroscience, Leipzig, Germany.
Felix Blankenburg Bernstein Center for Computational Neuroscience, Berlin.
John O'Doherty, California Institute of Technology, Pasadena, California, USA.
Christian Doeller Donders Institute, Radboud University Nijmegen, Netherlands.

References

[1] Kiebel, S. J., von Kriegstein, K., Daunizeau, J., & Friston, K. J. (2009). Recognizing sequences of sequences PLoS Comp Biol, (5), e1000464.
[2] Kiebel, S. J., Daunizeau, J., & Friston, K. J. (2008). A hierarchy of time-scales and the brain PLoS Computational Biology, 4(11), e1000209.
[3] Kiebel, S. J., Daunizeau, J., & Friston, K. J. (2009). Perception and hierarchical dynamics Frontiers in Neuroinformatics, (3:20).
[4] Daunizeau J, den Ouden HEM, Pessiglione M, Kiebel SJ, Friston KJ, et al. Observing the Observer (II): Deciding When to Decide. PLoS ONE 5(12): e15555
[5] von Kriegstein K., Dogan O., Grüter M., Giraud A. L., Kell C. A., Grüter T., Kleinschmidt A., Kiebel S. J. (2008). Simulation of talking faces in the human brain improves auditory speech recognition Proc Natl Acad Sci U S A, 105(18):6747-52

Footer



Last update: Oct 9, 2013 11.33.02 am
Copyright © 2013 Max Planck Institute for Human Cognitive and Brain Sciences