HISTORY . Summary of previous investigations.

    Laboratory have been  working in the field of neural networks for about  20  years.  The main  direction of investigations  have  been related  to modeling the  functional  behavior of  structures in the central  nervous system. The Laboratory was among the first to apply new  mathematical  tools -  multidimensional  interacting  Markovian processes  and  fields -  to the  analysis of  dynamical  regimes in stochastic neural networks (large-scale approach with many variables).

    Also the Laboratory was engaged in  a pioneer work  on developing numerical  algorithms  to  investigate  the  bifurcations  of steady states  and  limit  cycles  in  nonlinear  dynamical  systems  under parameter  variation   (small-scale  approach  with  several  main variables).  These investigations  showed  that  cooperative effects like  physical phase-transitions  and  synchronization phenomena are normal to occur in biological neural networks despite a very chaotic spike activity of single neurons.  In addition, metastable states of neural  networks  were  proved  to be  useful  to model  short- term memory  in a series  of theoretical  and simulation works . This led to development of a  neural  network  model  based  on  biologically plausible  constraints  on  the  elements  of a network  (similar to integrate-and-fire  neurons )  which  was  capable  of demonstrating metastable states.

    This basic neural network model  was used  to model low frequency oscillations  in the  septum,  the  habituation  in the  hippocampus, metastable  states  in the cortex  and  the cerebellum.