Signal transmission in multilayer asynchronous neural networks

It is believed that common input to nearby neurons leads to their synchronous spiking. However, recent studies have shown that recurrent neural networks can generate an asynchronous state characterized by low mean spiking correlations despite substantial amounts of shared input. The asynchronous state is generated by the interaction of excitatory and inhibitory populations, which is called active decorrelation. Here, we investigate the advantage of the active decorrelation on signal transmission in multilayer neural networks. The results of numerical simulations show that the active decorrelation is suitable for transmission of rate code because it can suppress the layer-by-layer growth of correlation.