Encoding information into autonomously bursting neural network with pairs of time-delayed pulses.

摘要: Biological neural networks with many plastic synaptic connections can store external input information in the map of weights as a form unsupervised learning. However, same network often produces dramatic reverberating events which neurons fire almost simultaneously – phenomenon coined ‘population burst.’ The autonomous bursting activity is consequence delicate balance between recurrent excitation and self-inhibition; such, any periodic sequences burst-generating stimuli delivered even at low frequency (~1 Hz) easily suppress entire connectivity. Here we demonstrate that ‘Δt paired-pulse stimulation’, be novel way for encoding spatially-distributed high-frequency (~10 Hz) into such system without causing complete suppression. encoded memory probed simply by delivering multiple probing pulses then estimating precision arrival times subsequent evoked bursts.