Constructing Precisely Computing Networks with Biophysical Spiking Neurons

摘要: While spike timing has been shown to carry detailed stimulus information at the sensory periphery, its possible role in network computation is less clear. Most models of by neural networks are based on population firing rates. In equivalent spiking implementations, assumed be random such that averaging across populations neurons recovers rate-based approach. Recently, however, Deneve and colleagues have suggested behavior may fundamental how neuronal compute, with precise determined each neuron9s contribution producing desired output (Boerlin Deneve, 2011; Boerlin et al., 2013). By postulating neuron fires reduce error network9s output, it was demonstrated linear computations can performed integrate-and-fire communicate through instantaneous synapses. This left open, possibility realistic networks, conductance-based subthreshold nonlinearity slower timescales biophysical synapses, not fit into this framework. Here, we show spike-based approach extended biophysically plausible networks. We then our reproduces a number key features cortical including irregular Poisson-like times tight balance between excitation inhibition. Lastly, discuss model scales size or “recorded” from larger computing network. These results significantly increase biological plausibility computation. SIGNIFICANCE STATEMENT derive standard spike-generating currents synapses computes upon principle important for irregular, times, computation, uncover several components work together efficiently out