Dual coding in a network of spiking neurons: aperiodic spikes and stable firing rates

Temporal coding as well as firing rate coding has been paid much attention as a possible way of information representation in the brain. The relation between the firing rate coding and the temporal coding, however, has been clarified neither theoretically nor experimentally. In this study, we propose a neural network model composed of spiking neurons, in which the spatio-temporal structure of spikes is chaotic but the spatial pattern of firing rates is nearly steady and stable. When Poisson-process asynchronous patterns are input to the model, spatio-temporal outputs of the neurons show aperiodic and chaotic patterns. On the other hand, we also observe stable spatial patterns of mean firing rates, which depend upon the sum of effective synaptic weights. These results suggest that the neural network model has two completely different coding mechanisms simultaneously. In other words, the model shows that temporal coding and firing rate coding can coexist. Each of these emergent properties can be expected to play a role in different coding, namely in different information processing in the brain