Specialisation in recurrent neural networks with spike-timing-dependent plasticity

Matthieu Gilson (1) , David B. Grayden (1) , Doreen A. Thomas (1) , Anthony N. Burkitt (1) , J. Leo van Hemmen (2)

1 Department of Electrical and Electronic Engineering [Melbourne]
2 Physik Department [Garching]

We examine in this paper how Spike-Timing-Dependent Plasticity (STDP) can implement the strengthening of recurrent excitatory connections in addition to their homeostatic equilibrium in a network of Poisson neurons stimulated by correlated pools of external inputs. This phenomenon is determined by the interplay between STDP, the neuronal mechanisms related to the post-synaptic response and the input correlation structure. Our results can be related to the emergence of functional areas in recurrently connected networks, i.e. self-organising maps.

Mots clés: Learning; Spiking neurons; Synaptic plasticity

Domaines: Neurosciences [q-bio.NC]; Neurosciences

Matthieu Gilson, David B. Grayden, Doreen A. Thomas, Anthony N. Burkitt, J. Leo van Hemmen. Specialisation in recurrent neural networks with spike-timing-dependent plasticity. Deuxième conférence française de Neurosciences Computationnelles, "Neurocomp08", Oct 2008, Marseille, France. ⟨hal-00331547⟩