Correlation of multiple neuronal spike trains using the back-propagation error correction algorithm

With the recent development of simultaneously-recorded multiple neuron recording techniques, such as voltage-sensitive dyes or multiple-electrode arrays, recordings from thousands of neurons are possible. Analysis of these simultaneously recorded spike-train data using conventional pair-wise correlation techniques typically requires combinatorially great numbers of correlograms and an immense amount of computing time. To alleviate these problems, we propose a parallel-distributed processing (PDP) network using the back-propagating error-correction algorithm to identify patterns of discharge activity from simultaneously recorded multiple spike trains.