View count: 179

Speakers

Prof. Ting-Kuo Lee 
Honorary Center Scientist, NCTS
Department of Physics, NSYSU
 
Statistical Modeling of Collective Behavior of Neuronal Activities in Mouse Brains 
Characterizing the neural activities in human or mouse brains and understanding its functions is one of the most challenging problems in science. The complex networks formed by vast numbers of neurons and their interconnections made the problem almost intractable. However the recent breakthrough in recording electrical activity of large number of mouse neurons provide a renewed interest of the statistical modeling of the collective behavior of these neurons. Based on a purely data-driven phenomenological approach, Bialek[1] and collaborators have been able to construct a pair-wise-coupled spin glass model to reproduce the correlation of the observed data without invoking a priori postulates and assumptions about the underlying structure or interactions of the networks. The unexpected result is that the observed correlation can only be reproduced at the critical temperature of the constructed model. We will also show our analysis of the observed in-vivo recording data by a miniscope mounted on a free-moving mouse from our collaborators, Professor D.C. Wu’s group, in Taiwan. Although the experimental setup is very different from Bialek’s, result of these more than 10 sets of data shows good agreement.
Furthermore we also found the surprising result that the collective behavior of the neuron networks we had is close to the critical state of the calculated spin glass model. In fact, the neural network represented by this spin glass model has similar special features.  Details of the calculation techniques we used , the analysis and the implication of these features will be discussed.  
 
[1] E. Schneidman et al., Nature 440, 1007 (2006), L. Meshulam et al., Neuron 96, 1178 (2017).