Validation of long-term changes of evoked response with self-orgnization map

In order to analyze stability of stimulation evoked neuronal activity, it is necessary to analyze long-term changes of evoked response pattern induced by electrical stimulation. For this purpose, we use Self-Organization-Map (SOM), which continuously performs non-teacher learning. Neural network was alternately stimulated by two different electrodes. SOM maps to response pattern induced by each electrical stimulation in 30 × 30-2D space. Only at the first stage of the learning, SOM is forced to select a specific winner nodes previously assigned in order to associate a specific node to the input vector. We call the process "seeding". After seeding procedure, the winner units correspond to the response patterns induced by two different inputs are separately mapped. We confirmed that response patterns of electrical stimulation were stimulation specific and they were almost stable. In addition, Not small part of spontaneous activities share the activity patterns with evoked responses, suggesting that inputs were not tightly coupled to their response patterns.