Structural plasticity in recurrent cortical networks

Abstract

Event Abstract Back to Event Structural plasticity in recurrent cortical networks Moritz Deger1*, Moritz Helias1, Markus Diesmann2 and Stefan Rotter1 1 Albert-Ludwigs-University, Germany 2 RIKEN Brain Science Institute, Japan We study recurrent neural networks exhibiting structural plasticity by formation and elimination of synapses: synaptic death is controlled by a biologically realistic correlation dependent learning rule [1] and synapse formation takes place in a random manner. The interplay between the network dynamics and the correlation dependent evolution of the network structure exhibits an interesting feature: We observe emergence of cell assemblies in initially random cortical networks upon correlated stimulation of a subgroup of neurons. To gain a quantitative understanding, we reduce the detailed spike based learning dynamics to effective rate equations where the network dynamics follows the structure adiabatically. We show that this description captures the essential features of the interplay between structural plasticity and network dynamics. Comparison to direct numerical simulations proves the approximate procedures applied on many levels adequate.