English
 
Help Privacy Policy Disclaimer
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Poster

Representation of spatiotemporal sequences in hippocampus: A dynamic Bayesian model

MPS-Authors
/persons/resource/persons19554

Bitzer,  Sebastian
Department Neurology, MPI for Human Cognitive and Brain Sciences, Max Planck Society;

Aponte,  Eduardo
Department Neurology, MPI for Human Cognitive and Brain Sciences, Max Planck Society;

/persons/resource/persons19770

Kiebel,  Stefan J.
Department Neurology, MPI for Human Cognitive and Brain Sciences, Max Planck Society;

External Resource
No external resources are shared
Fulltext (public)
There are no public fulltexts stored in PuRe
Supplementary Material (public)
There is no public supplementary material available
Citation

Bitzer, S., Aponte, E., & Kiebel, S. J. (2011). Representation of spatiotemporal sequences in hippocampus: A dynamic Bayesian model. Poster presented at BC11: Computational Neuroscience & Neurotechnology Bernstein Conference & Neurex Annual Meeting 2011, Freiburg im Breisgau, Germany.


Cite as: http://hdl.handle.net/11858/00-001M-0000-0012-19E6-B
Abstract
Place cells in the hippocampus of rodents fire preferentially when the animal is in a specific location of its environment. Sequential firing of place cells has been observed not only during movement through a sequence of locations but also during rest periods [1]. This means that place cells learnt to represent experienced sequences of places, i.e. paths. Although several models have been proposed for how individual place cells come to represent specific locations, there are relatively few models for learning and recognition of paths by hippocampal neurons or populations. Here, we present a probabilistic model of population coding in hippocampus based on Lotka-Volterra dynamics [2], which replicates a number of experimental key findings. The model provides for a simple mechanism of how paths can be learnt and encoded as sequential activation of place cells. In the present approach we assume that dynamic templates are encoded by recurrently connected populations of neurons where learning maps these templates to the actually experienced spatiotemporal input [3], i.e. the sensory dynamics induced by moving along a path. After learning, the model implements a predictive coding scheme based on Bayesian inference for nonlinear dynamical systems [4] to recognize paths based on ongoing sensory input. In other words, the model recognizes and predicts the current path by anticipating the sequential activation of place cells. In simulations we show that this recognition and prediction scheme is robust against minor deviations from the learnt path, noise in place cells activity and the speed with which the animal moves along a path. Furthermore, once a path is learnt, the model can also recognize and predict the reverse path, which mirrors another experimental key finding [1]. In summary, we present a probabilistic model of sequential activations of place cells which qualitatively explains a wide range of experimental findings and may be the basis for further research into the role of sequential firing in hippocampus.