Neural Network Analysis Of Neuronal Spike-trains

Spike-train tomographic scanning (ST-scanning) is a feature extraction method, useful in the separation of component action potentials from composite spike-trains. Using ST-scanning, spike-train components are differentially represented in histograms according to the scan angle used. These histogram waveforms are then analyzed using a 3-layer Perceptron, employing an LMS learning algorithm. The neural network is first trained to relate specific ST-scan histograms to known spike-train patterns. After training, the Perceptron is run in feed-forward mode to allow the separation of component action potentials from composite waveforms.