[Comp-neuro] PhD-Position "Higher-order correlations among spiking neurons induced by network structure " at the Bernstein Center Freiburg, Erasmus Mundus Joint Doctoral Program "EuroSPIN"

Janina Kirsch kirsch at bcf.uni-freiburg.de
Fri Nov 25 16:03:13 CET 2011


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%                                                                              %%
%% Higher-order correlations among spiking neurons induced by network structure %%
%%                                                                              %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Supervisors: Prof. Dr. Stefan Rotter (Bernstein Center Freiburg, Germany)
	       Prof. Dr. John Hertz (NORDITA, Stockholm, Sweden)
	       Prof. Dr. Erik Aurell (KTH, Stockholm, Sweden)

Aim of the project:
Nerve cells are highly sensitive to synchronous input from larger groups of neurons [1]. Which synchronous patterns are favored by a recurrent network, therefore, depends to a large degree on network structure. Recently, we were able to dissect the contribution of specific structural motifs in networks of arbitrary topology to pairwise correlations [3], based on a minimal model for networks of spiking neurons [2]. The case of higher-order correlations, however, is complicated by the fact that several different concepts to describe multi-neuron interactions are in use. The so-called log-linear model represents a generalization of the well-studied Ising model [4,5] and therefore permits one to exploit techniques from statistical physics. Models based on stochastic point processes, on the other hand, have frequently been employed as generative models in neuroscience [1,6]. These represent a suitable starting point to develop efficient methods for neuronal data analysis, because they admit a natural link to multivariate cumulants [6]. In this project, we strive to generalize our dynamical systems approach to build a concrete physical interpretation of higher-order correlations and find out which network motifs are responsible for their generation. Recent methodological developments in neuroanatomy that allow one to assess the microstructure of brain networks at an unprecedented level (“connectomics”), in fact, generate a demand for a theory of this sort.

1.	Kuhn A, Aertsen A, Rotter S. Higher-order statistics of input ensembles and the response of simple model neurons. Neural Computation 15(1): 67-101, 2003
2.	Hawkes AG. Point spectra of some mutually exciting point processes. J R Stat Soc Series B Methodol 33: 438-443, 1971
3.	Pernice V, Staude B, Cardanobile S, Rotter S. How Structure Determines Correlations in Neuronal Networks. PLoS Computational Biology 7(5): e1002059, 2011
4.	Roudi Y, Aurell E, Hertz JA. Statistical physics of pairwise probability models. Front Comput Neurosci 3: 22, 2009
5.	Roudi Y, Hertz J. Mean field theory for nonequilibrium network reconstruction. Phys Rev Lett 106(4): 048702, 2011
6.	Staude B, Grün S, Rotter S. Higher-order correlations and cumulants. In: Grün S, Rotter S (eds) Analysis of Parallel Spike Trains. Springer Series in Computational Neuroscience, Volume 7, 2010

Profile of the ideal candidate:
We seek a statistical physicist, mathematician or computational neuroscientist who is able to perform creative research along the lines described above. The ability to interact successfully with colleagues from other disciplines, in particular from neuroanatomy and neurophysiology, will be necessary. He/she will be enrolled in the PhD program in Computational Neuroscience at the BCF (http://www.bcf.uni-freiburg.de/teaching-and-training/phd-program). We plan a joint supervision by Stefan Rotter (Freiburg), John Hertz (Copenhagen & Stockholm) and Erik Aurell (Stockholm).

Application deadline is November 30, 2011

Please apply here: www.kth.se/eurospin 




More information about the Comp-neuro mailing list