Spike-timing dependent plasticity in recurrently connected networks with fixed external inputs

This paper investigates spike-timing dependent plasticity (STDP) for recurrently connected weights in a network with fixed external inputs (homogeneous Poisson pulse trains). We use a dynamical system to model the network activity and predict its asymptotic evolution, which turns out to qualitatively depend on the learning parameters and the correlation structure of the inputs. Our predictions are supported by numerical simulations of Poisson neuron networks in general cases as well as for certain cases when using Integrate-And-Fire (IF) neurons.