Self-Organizing Network Learning of Sub-Millisecond Temporal Coded Information

In this article, we report a simulation result of unsupervised network learning characterized as temporally and spatially local. After the learning, the network preserves an input sequence whose intervals vary in sub-millisecond order, which is much smaller than the spike emission interval of the neurons. This is achieved by the forming of systematic local structures. This formation is done by selecting appropriate connections from the input neurons and other processing neurons. As a result, the network successfully becomes a sub-millisecond temporal information processing system in a self-organizing manner.